From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 42670C433EF for ; Wed, 9 Mar 2022 09:29:05 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S232091AbiCIJaB (ORCPT ); Wed, 9 Mar 2022 04:30:01 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:53394 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229623AbiCIJ36 (ORCPT ); Wed, 9 Mar 2022 04:29:58 -0500 Received: from mga11.intel.com (mga11.intel.com [192.55.52.93]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id DE597EB30C for ; Wed, 9 Mar 2022 01:29:00 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1646818140; x=1678354140; h=from:to:cc:subject:references:date:in-reply-to: message-id:mime-version; bh=CVn+ZT5avv8HOvgaAaNLsjJaTQhv2nlshh1mq8i1iV0=; b=PtMsaIwJ3kfOyZcsWNiS4AJcVsSadbaZ5ww5MZj22Ng9GedneYnMidbt H64qXkFfc4ATBMgaAbcN7tHruJLZDPLdnMwjN2CkEGZ63AOeCP7s21QFL 6bFMiL0yqo3XIk+Ah1WN7RjpIPPUFPVFu9WCesumimVRXZ52ZeiPSxS6v P6cSwLiVEcNV5x7u/9u8veSUQOlISVz9noGAYzeudkJh+A2HcYe3BD7vo T45PVQ1LxE6wJktGiOgEj03kACyaTwhOMryqRT6MiY51QjhVo0qp0tF9v hghqEyUtiixA56xxuCX02+g2F0oxrf9BfSDcXsyv1fALKqYNB4d9qcdlc w==; X-IronPort-AV: E=McAfee;i="6200,9189,10280"; a="252501840" X-IronPort-AV: E=Sophos;i="5.90,167,1643702400"; d="scan'208";a="252501840" Received: from orsmga007.jf.intel.com ([10.7.209.58]) by fmsmga102.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 09 Mar 2022 01:29:00 -0800 X-IronPort-AV: E=Sophos;i="5.90,167,1643702400"; d="scan'208";a="537936493" Received: from yhuang6-desk2.sh.intel.com (HELO yhuang6-desk2.ccr.corp.intel.com) ([10.239.13.94]) by orsmga007-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 09 Mar 2022 01:28:57 -0800 From: "Huang, Ying" To: Peter Zijlstra , kernel test robot , Mel Gorman Cc: LKML , x86@kernel.org, lkp@lists.01.org, lkp@intel.com, fengwei.yin@intel.com, aubrey.li@linux.intel.com, yu.c.chen@intel.com Subject: Re: [LKP] Re: [sched/numa] 0fb3978b0a: stress-ng.fstat.ops_per_sec -18.9% regression References: <20220302140220.GA32817@xsang-OptiPlex-9020> <87mti7s9xz.fsf@yhuang6-desk2.ccr.corp.intel.com> Date: Wed, 09 Mar 2022 17:28:55 +0800 In-Reply-To: <87mti7s9xz.fsf@yhuang6-desk2.ccr.corp.intel.com> (Ying Huang's message of "Thu, 03 Mar 2022 16:43:20 +0800") Message-ID: <87tuc7fp9k.fsf@yhuang6-desk2.ccr.corp.intel.com> User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/27.1 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=ascii Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, All, "Huang, Ying" writes: > Hi, Oliver, > > Thanks for report. > > I still cannot connect the regression with the patch yet. To double > check, I have run test again with "sched_verbose" kernel command line, > and verified that the sched_domain isn't changed at all with the patch. > > kernel test robot writes: >> 0.11 6% +0.1 0.16 4% perf-profile.self.cycles-pp.update_rq_clock >> 0.00 +0.1 0.06 6% perf-profile.self.cycles-pp.memset_erms >> 0.00 +0.1 0.07 5% perf-profile.self.cycles-pp.get_pid_task >> 0.06 7% +0.1 0.17 6% perf-profile.self.cycles-pp.select_task_rq_fair >> 0.54 5% +0.1 0.68 perf-profile.self.cycles-pp.lockref_put_return >> 4.26 +1.1 5.33 perf-profile.self.cycles-pp.common_perm_cond >> 15.45 +4.9 20.37 perf-profile.self.cycles-pp.lockref_put_or_lock >> 20.12 +6.7 26.82 perf-profile.self.cycles-pp.lockref_get_not_dead > > From the perf-profile above, the most visible change is more cycles in > lockref_get_not_dead(), which will loop with cmpxchg on > dentry->d_lockref. So this appears to be related to the memory layout. > I will try to debug that. > > Because stress-ng is a weird "benchmark" although it's a very good > functionality test, and I cannot connect the patch with the test case > and performance metrics collected. I think this regression should be a > low priority one which shouldn't prevent the merging etc. But I will > continue to investigate the regression to try to root cause it. Done more investigation for this. It turns out the sched_domain has been changed after commit 0fb3978b0a, although it's not shown in default sched_verbose output. sd->imb_numa_nr of level "NUMA" has been changed from 24 to 12 after the commit. So the following debug patch restore the performance. >From 6c79c858248196ea71da4be02a0774bb6e277581 Mon Sep 17 00:00:00 2001 From: Huang Ying Date: Wed, 9 Mar 2022 10:45:18 +0800 Subject: [PATCH] dbg: restore sd->imb_numa_nr --- kernel/sched/topology.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c index 5e05fb7a817b..c348747e4166 100644 --- a/kernel/sched/topology.c +++ b/kernel/sched/topology.c @@ -2351,6 +2351,10 @@ build_sched_domains(const struct cpumask *cpu_map, struct sched_domain_attr *att sd->name, imb, factor, sd->imb_numa_nr); #endif } +#ifdef CONFIG_SCHED_DEBUG + if (!strcmp(sd->name, "NUMA")) + sd->imb_numa_nr = 24; +#endif } } -- 2.30.2 I haven't researched much about the influence of sd->imb_numa_nr. But it seems that commit 0fb3978b0a just follows the behavior of a normal 2 sockets (2 LLCs) machine. This is intended. So the regression report isn't important for the commit itself. About the impact of sd->imb_numa_nr, we will continue work on that. Best Regards, Huang, Ying