From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1422920AbeCBIaV (ORCPT ); Fri, 2 Mar 2018 03:30:21 -0500 Received: from mga02.intel.com ([134.134.136.20]:42242 "EHLO mga02.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1422857AbeCBIaU (ORCPT ); Fri, 2 Mar 2018 03:30:20 -0500 X-Amp-Result: UNSCANNABLE X-Amp-File-Uploaded: False X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.47,411,1515484800"; d="scan'208";a="34088912" Date: Fri, 2 Mar 2018 16:31:19 +0800 From: Aaron Lu To: Michal Hocko Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, Andrew Morton , Huang Ying , Dave Hansen , Kemi Wang , Tim Chen , Andi Kleen , Vlastimil Babka , Mel Gorman , Matthew Wilcox , David Rientjes Subject: Re: [PATCH v4 3/3] mm/free_pcppages_bulk: prefetch buddy while not holding lock Message-ID: <20180302083119.GD6356@intel.com> References: <20180301062845.26038-1-aaron.lu@intel.com> <20180301062845.26038-4-aaron.lu@intel.com> <20180301140044.GK15057@dhcp22.suse.cz> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20180301140044.GK15057@dhcp22.suse.cz> User-Agent: Mutt/1.9.2 (2017-12-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Mar 01, 2018 at 03:00:44PM +0100, Michal Hocko wrote: > On Thu 01-03-18 14:28:45, Aaron Lu wrote: > > When a page is freed back to the global pool, its buddy will be checked > > to see if it's possible to do a merge. This requires accessing buddy's > > page structure and that access could take a long time if it's cache cold. > > > > This patch adds a prefetch to the to-be-freed page's buddy outside of > > zone->lock in hope of accessing buddy's page structure later under > > zone->lock will be faster. Since we *always* do buddy merging and check > > an order-0 page's buddy to try to merge it when it goes into the main > > allocator, the cacheline will always come in, i.e. the prefetched data > > will never be unused. > > > > In the meantime, there are two concerns: > > 1 the prefetch could potentially evict existing cachelines, especially > > for L1D cache since it is not huge; > > 2 there is some additional instruction overhead, namely calculating > > buddy pfn twice. > > > > For 1, it's hard to say, this microbenchmark though shows good result but > > the actual benefit of this patch will be workload/CPU dependant; > > For 2, since the calculation is a XOR on two local variables, it's expected > > in many cases that cycles spent will be offset by reduced memory latency > > later. This is especially true for NUMA machines where multiple CPUs are > > contending on zone->lock and the most time consuming part under zone->lock > > is the wait of 'struct page' cacheline of the to-be-freed pages and their > > buddies. > > > > Test with will-it-scale/page_fault1 full load: > > > > kernel Broadwell(2S) Skylake(2S) Broadwell(4S) Skylake(4S) > > v4.16-rc2+ 9034215 7971818 13667135 15677465 > > patch2/3 9536374 +5.6% 8314710 +4.3% 14070408 +3.0% 16675866 +6.4% > > this patch 10338868 +8.4% 8544477 +2.8% 14839808 +5.5% 17155464 +2.9% > > Note: this patch's performance improvement percent is against patch2/3. > > I am really surprised that this has such a big impact. Is this a win on > other architectures as well? For NUMA machines, I guess so. But I didn't test other archs so can't say for sure. > > > [changelog stole from Dave Hansen and Mel Gorman's comments] > > https://lkml.org/lkml/2018/1/24/551 > > Please use http://lkml.kernel.org/r/ for references because > lkml.org is quite unstable. It would be > http://lkml.kernel.org/r/148a42d8-8306-2f2f-7f7c-86bc118f8ccd@intel.com > here. Good to know this, thanks! From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-pg0-f69.google.com (mail-pg0-f69.google.com [74.125.83.69]) by kanga.kvack.org (Postfix) with ESMTP id DB92D6B0007 for ; Fri, 2 Mar 2018 03:30:20 -0500 (EST) Received: by mail-pg0-f69.google.com with SMTP id m19so3859733pgv.5 for ; Fri, 02 Mar 2018 00:30:20 -0800 (PST) Received: from mga07.intel.com (mga07.intel.com. [134.134.136.100]) by mx.google.com with ESMTPS id u25si4479139pfm.164.2018.03.02.00.30.19 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Fri, 02 Mar 2018 00:30:19 -0800 (PST) Date: Fri, 2 Mar 2018 16:31:19 +0800 From: Aaron Lu Subject: Re: [PATCH v4 3/3] mm/free_pcppages_bulk: prefetch buddy while not holding lock Message-ID: <20180302083119.GD6356@intel.com> References: <20180301062845.26038-1-aaron.lu@intel.com> <20180301062845.26038-4-aaron.lu@intel.com> <20180301140044.GK15057@dhcp22.suse.cz> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20180301140044.GK15057@dhcp22.suse.cz> Sender: owner-linux-mm@kvack.org List-ID: To: Michal Hocko Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, Andrew Morton , Huang Ying , Dave Hansen , Kemi Wang , Tim Chen , Andi Kleen , Vlastimil Babka , Mel Gorman , Matthew Wilcox , David Rientjes On Thu, Mar 01, 2018 at 03:00:44PM +0100, Michal Hocko wrote: > On Thu 01-03-18 14:28:45, Aaron Lu wrote: > > When a page is freed back to the global pool, its buddy will be checked > > to see if it's possible to do a merge. This requires accessing buddy's > > page structure and that access could take a long time if it's cache cold. > > > > This patch adds a prefetch to the to-be-freed page's buddy outside of > > zone->lock in hope of accessing buddy's page structure later under > > zone->lock will be faster. Since we *always* do buddy merging and check > > an order-0 page's buddy to try to merge it when it goes into the main > > allocator, the cacheline will always come in, i.e. the prefetched data > > will never be unused. > > > > In the meantime, there are two concerns: > > 1 the prefetch could potentially evict existing cachelines, especially > > for L1D cache since it is not huge; > > 2 there is some additional instruction overhead, namely calculating > > buddy pfn twice. > > > > For 1, it's hard to say, this microbenchmark though shows good result but > > the actual benefit of this patch will be workload/CPU dependant; > > For 2, since the calculation is a XOR on two local variables, it's expected > > in many cases that cycles spent will be offset by reduced memory latency > > later. This is especially true for NUMA machines where multiple CPUs are > > contending on zone->lock and the most time consuming part under zone->lock > > is the wait of 'struct page' cacheline of the to-be-freed pages and their > > buddies. > > > > Test with will-it-scale/page_fault1 full load: > > > > kernel Broadwell(2S) Skylake(2S) Broadwell(4S) Skylake(4S) > > v4.16-rc2+ 9034215 7971818 13667135 15677465 > > patch2/3 9536374 +5.6% 8314710 +4.3% 14070408 +3.0% 16675866 +6.4% > > this patch 10338868 +8.4% 8544477 +2.8% 14839808 +5.5% 17155464 +2.9% > > Note: this patch's performance improvement percent is against patch2/3. > > I am really surprised that this has such a big impact. Is this a win on > other architectures as well? For NUMA machines, I guess so. But I didn't test other archs so can't say for sure. > > > [changelog stole from Dave Hansen and Mel Gorman's comments] > > https://lkml.org/lkml/2018/1/24/551 > > Please use http://lkml.kernel.org/r/ for references because > lkml.org is quite unstable. It would be > http://lkml.kernel.org/r/148a42d8-8306-2f2f-7f7c-86bc118f8ccd@intel.com > here. Good to know this, thanks! -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org