From patchwork Tue Dec 8 15:35:00 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Mel Gorman X-Patchwork-Id: 11958813 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-16.8 required=3.0 tests=BAYES_00,DKIMWL_WL_HIGH, DKIM_SIGNED,DKIM_VALID,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER, INCLUDES_PATCH,MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,URIBL_BLOCKED, USER_AGENT_GIT autolearn=unavailable autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 5DC6BC4361B for ; Tue, 8 Dec 2020 15:36:58 +0000 (UTC) Received: from merlin.infradead.org (merlin.infradead.org [205.233.59.134]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 1F73323A7B for ; Tue, 8 Dec 2020 15:36:58 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 1F73323A7B Authentication-Results: mail.kernel.org; dmarc=none (p=none dis=none) header.from=techsingularity.net Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=linux-arm-kernel-bounces+linux-arm-kernel=archiver.kernel.org@lists.infradead.org DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=lists.infradead.org; s=merlin.20170209; h=Sender:Content-Transfer-Encoding: Content-Type:Cc:List-Subscribe:List-Help:List-Post:List-Archive: List-Unsubscribe:List-Id:MIME-Version:References:In-Reply-To:Message-Id:Date: Subject:To:From:Reply-To:Content-ID:Content-Description:Resent-Date: Resent-From:Resent-Sender:Resent-To:Resent-Cc:Resent-Message-ID:List-Owner; bh=vOS2BcPFyTJNy5MNjJm1quNp5SS5NlIW+vNab9cCLH4=; b=RQbf5fmuVPlzvRbiKt0xCB3A2 wEhIUBZ/QLRnBQVWMyOpJQmv+j8I16xL5QmpaYDpnONdY7bXUaPlFRyzxYgOd28qeHHsugBxW+3tR PjDvW4EiJRkvAFgWPj++6F2j+RAYvfrvIaR/nyabCV4f8yx1G5a11WsE0iCxLC30C9zOxlthgPlwI rAJnkCCERJS2n7YZ1wGthbgkuYHIA4mRGXXsdLj9G1tybE2A7UN4hRLm/5eOxbwtvIWX7GxESoDe5 OPK+kfZ+uKpJV4JcN4q+cYI7UpM6VeM1s2jGNxRvV+ApAmJfRpHkhNJpmCbnwbFgU4EY610nYucXV GzRm4gnjQ==; Received: from localhost ([::1] helo=merlin.infradead.org) by merlin.infradead.org with esmtp (Exim 4.92.3 #3 (Red Hat Linux)) id 1kmf1o-0006mu-Dx; Tue, 08 Dec 2020 15:35:48 +0000 Received: from outbound-smtp38.blacknight.com ([46.22.139.221]) by merlin.infradead.org with esmtps (Exim 4.92.3 #3 (Red Hat Linux)) id 1kmf1l-0006lz-QP for linux-arm-kernel@lists.infradead.org; Tue, 08 Dec 2020 15:35:46 +0000 Received: from mail.blacknight.com (pemlinmail04.blacknight.ie [81.17.254.17]) by outbound-smtp38.blacknight.com (Postfix) with ESMTPS id 347DD19D6 for ; Tue, 8 Dec 2020 15:35:44 +0000 (GMT) Received: (qmail 12810 invoked from network); 8 Dec 2020 15:35:44 -0000 Received: from unknown (HELO stampy.112glenside.lan) (mgorman@techsingularity.net@[84.203.22.4]) by 81.17.254.9 with ESMTPA; 8 Dec 2020 15:35:43 -0000 From: Mel Gorman To: Peter Ziljstra , Ingo Molnar , LKML Subject: [PATCH 3/4] sched/fair: Do not replace recent_used_cpu with the new target Date: Tue, 8 Dec 2020 15:35:00 +0000 Message-Id: <20201208153501.1467-4-mgorman@techsingularity.net> X-Mailer: git-send-email 2.26.2 In-Reply-To: <20201208153501.1467-1-mgorman@techsingularity.net> References: <20201208153501.1467-1-mgorman@techsingularity.net> MIME-Version: 1.0 X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.8.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20201208_103546_047193_55D3367D X-CRM114-Status: GOOD ( 17.41 ) X-BeenThere: linux-arm-kernel@lists.infradead.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Barry Song , Juri Lelli , Vincent Guittot , Aubrey Li , Mel Gorman , Valentin Schneider , Linux-ARM Sender: "linux-arm-kernel" Errors-To: linux-arm-kernel-bounces+linux-arm-kernel=archiver.kernel.org@lists.infradead.org After select_idle_sibling, p->recent_used_cpu is set to the new target. However on the next wakeup, prev will be the same as recent_used_cpu unless the load balancer has moved the task since the last wakeup. It still works, but is less efficient than it can be after all the changes that went in since that reduce unnecessary migrations, load balancer changes etc. This patch preserves recent_used_cpu for longer. With tbench on a 2-socket CascadeLake machine, 80 logical CPUs, HT enabled 5.10.0-rc6 5.10.0-rc6 baseline-v2 altrecent-v2 Hmean 1 508.39 ( 0.00%) 502.05 * -1.25%* Hmean 2 986.70 ( 0.00%) 983.65 * -0.31%* Hmean 4 1914.55 ( 0.00%) 1920.24 * 0.30%* Hmean 8 3702.37 ( 0.00%) 3663.96 * -1.04%* Hmean 16 6573.11 ( 0.00%) 6545.58 * -0.42%* Hmean 32 10142.57 ( 0.00%) 10253.73 * 1.10%* Hmean 64 14348.40 ( 0.00%) 12506.31 * -12.84%* Hmean 128 21842.59 ( 0.00%) 21967.13 * 0.57%* Hmean 256 20813.75 ( 0.00%) 21534.52 * 3.46%* Hmean 320 20684.33 ( 0.00%) 21070.14 * 1.87%* The different was marginal except for 64 threads which showed in the baseline that the result was very unstable where as the patch was much more stable. This is somewhat machine specific as on a separate 80-cpu Broadwell machine the same test reported. 5.10.0-rc6 5.10.0-rc6 baseline-v2 altrecent-v2 Hmean 1 310.36 ( 0.00%) 291.81 * -5.98%* Hmean 2 340.86 ( 0.00%) 547.22 * 60.54%* Hmean 4 912.29 ( 0.00%) 1063.21 * 16.54%* Hmean 8 2116.40 ( 0.00%) 2103.60 * -0.60%* Hmean 16 4232.90 ( 0.00%) 4362.92 * 3.07%* Hmean 32 8442.03 ( 0.00%) 8642.10 * 2.37%* Hmean 64 11733.91 ( 0.00%) 11473.66 * -2.22%* Hmean 128 17727.24 ( 0.00%) 16784.23 * -5.32%* Hmean 256 16089.23 ( 0.00%) 16110.79 * 0.13%* Hmean 320 15992.60 ( 0.00%) 16071.64 * 0.49%* schedstats were not used in this series but from an earlier debugging effort, the schedstats after the test run were as follows; Ops SIS Search 5653107942.00 5726545742.00 Ops SIS Domain Search 3365067916.00 3319768543.00 Ops SIS Scanned 112173512543.00 99194352541.00 Ops SIS Domain Scanned 109885472517.00 96787575342.00 Ops SIS Failures 2923185114.00 2950166441.00 Ops SIS Recent Used Hit 56547.00 118064916.00 Ops SIS Recent Used Miss 1590899250.00 354942791.00 Ops SIS Recent Attempts 1590955797.00 473007707.00 Ops SIS Search Efficiency 5.04 5.77 Ops SIS Domain Search Eff 3.06 3.43 Ops SIS Fast Success Rate 40.47 42.03 Ops SIS Success Rate 48.29 48.48 Ops SIS Recent Success Rate 0.00 24.96 First interesting point is the ridiculous number of times runqueues are enabled -- almost 97 billion times over the course of 40 minutes With the patch, "Recent Used Hit" is over 2000 times more likely to succeed. The failure rate also increases by quite a lot but the cost is marginal even if the "Fast Success Rate" only increases by 2% overall. What cannot be observed from these stats is where the biggest impact as these stats cover low utilisation to over saturation. If graphed over time, the graphs show that the sched domain is only scanned at negligible rates until the machine is fully busy. With low utilisation, the "Fast Success Rate" is almost 100% until the machine is fully busy. For 320 clients, the success rate is close to 0% which is unsurprising. Signed-off-by: Mel Gorman Reviewed-by: Vincent Guittot --- kernel/sched/fair.c | 9 +-------- 1 file changed, 1 insertion(+), 8 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 5c41875aec23..413d895bbbf8 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -6277,17 +6277,13 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target) /* Check a recently used CPU as a potential idle candidate: */ recent_used_cpu = p->recent_used_cpu; + p->recent_used_cpu = prev; if (recent_used_cpu != prev && recent_used_cpu != target && cpus_share_cache(recent_used_cpu, target) && (available_idle_cpu(recent_used_cpu) || sched_idle_cpu(recent_used_cpu)) && cpumask_test_cpu(p->recent_used_cpu, p->cpus_ptr) && asym_fits_capacity(task_util, recent_used_cpu)) { - /* - * Replace recent_used_cpu with prev as it is a potential - * candidate for the next wake: - */ - p->recent_used_cpu = prev; return recent_used_cpu; } @@ -6768,9 +6764,6 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags) } else if (wake_flags & WF_TTWU) { /* XXX always ? */ /* Fast path */ new_cpu = select_idle_sibling(p, prev_cpu, new_cpu); - - if (want_affine) - current->recent_used_cpu = cpu; } rcu_read_unlock();