From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754799AbdCIBtv (ORCPT ); Wed, 8 Mar 2017 20:49:51 -0500 Received: from mga14.intel.com ([192.55.52.115]:39058 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754236AbdCIBtt (ORCPT ); Wed, 8 Mar 2017 20:49:49 -0500 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.36,266,1486454400"; d="scan'208";a="74496015" From: "Huang\, Ying" To: Matthew Wilcox Cc: "Huang\, Ying" , Andrew Morton , , , Hugh Dickins , Shaohua Li , Minchan Kim , Rik van Riel Subject: Re: [PATCH -mm -v6 1/9] mm, swap: Make swap cluster size same of THP size on x86_64 References: <20170308072613.17634-1-ying.huang@intel.com> <20170308072613.17634-2-ying.huang@intel.com> <20170308125631.GX16328@bombadil.infradead.org> Date: Thu, 09 Mar 2017 09:45:11 +0800 In-Reply-To: <20170308125631.GX16328@bombadil.infradead.org> (Matthew Wilcox's message of "Wed, 8 Mar 2017 04:56:31 -0800") Message-ID: <87k27zi5p4.fsf@yhuang-dev.intel.com> User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/25.1 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=ascii Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, Matthew, Matthew Wilcox writes: > On Wed, Mar 08, 2017 at 03:26:05PM +0800, Huang, Ying wrote: >> In this patch, the size of the swap cluster is changed to that of the >> THP (Transparent Huge Page) on x86_64 architecture (512). This is for >> the THP swap support on x86_64. Where one swap cluster will be used to >> hold the contents of each THP swapped out. And some information of the >> swapped out THP (such as compound map count) will be recorded in the >> swap_cluster_info data structure. >> >> For other architectures which want THP swap support, >> ARCH_USES_THP_SWAP_CLUSTER need to be selected in the Kconfig file for >> the architecture. >> >> In effect, this will enlarge swap cluster size by 2 times on x86_64. >> Which may make it harder to find a free cluster when the swap space >> becomes fragmented. So that, this may reduce the continuous swap space >> allocation and sequential write in theory. The performance test in 0day >> shows no regressions caused by this. > > Well ... if there are no regressions found, why not change it > unconditionally? The value '256' seems relatively arbitrary (I bet it > was tuned by some doofus with a 486, 8MB RAM and ST506 hard drive ... > it certainly hasn't changed since git started in 2005) > > Might be worth checking with the PowerPC people to see if their larger > pages causes this smaller patch to perform badly: I found the huge page size is large not only on PowerPC, for example, on MIPS, the PMD_SHIFT could be from 21 to 29, depends on configuration. I don't know the situation for the other architectures. So I thought it may be better to let the architecture developers to determine whether to make the change and under which configuration. > diff --git a/mm/swapfile.c b/mm/swapfile.c > --- a/mm/swapfile.c > +++ b/mm/swapfile.c > @@ -199,7 +199,7 @@ static void discard_swap_cluster(struct swap_info_struct *si, > } > } > > -#define SWAPFILE_CLUSTER 256 > +#define SWAPFILE_CLUSTER HPAGE_PMD_NR > #define LATENCY_LIMIT 256 > > static inline void cluster_set_flag(struct swap_cluster_info *info, Best Regards, Huang, Ying