From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-wr0-f197.google.com (mail-wr0-f197.google.com [209.85.128.197]) by kanga.kvack.org (Postfix) with ESMTP id 5A7FB280271 for ; Fri, 5 Jan 2018 04:33:04 -0500 (EST) Received: by mail-wr0-f197.google.com with SMTP id s105so2304660wrc.23 for ; Fri, 05 Jan 2018 01:33:04 -0800 (PST) Received: from mx2.suse.de (mx2.suse.de. [195.135.220.15]) by mx.google.com with ESMTPS id f10si3396278wmi.275.2018.01.05.01.33.02 for (version=TLS1 cipher=AES128-SHA bits=128/128); Fri, 05 Jan 2018 01:33:03 -0800 (PST) Date: Fri, 5 Jan 2018 10:33:01 +0100 From: Michal Hocko Subject: Re: [PATCH 1/3] mm, numa: rework do_pages_move Message-ID: <20180105093301.GK2801@dhcp22.suse.cz> References: <20180103082555.14592-1-mhocko@kernel.org> <20180103082555.14592-2-mhocko@kernel.org> <20180105091443.GJ2801@dhcp22.suse.cz> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Sender: owner-linux-mm@kvack.org List-ID: To: Anshuman Khandual Cc: Andrew Morton , Zi Yan , Naoya Horiguchi , "Kirill A. Shutemov" , Vlastimil Babka , Andrea Reale , linux-mm@kvack.org, LKML On Fri 05-01-18 14:50:04, Anshuman Khandual wrote: > On 01/05/2018 02:44 PM, Michal Hocko wrote: > > On Fri 05-01-18 09:22:22, Anshuman Khandual wrote: > > [...] > >> Hi Michal, > >> > >> After slightly modifying your test case (like fixing the page size for > >> powerpc and just doing simple migration from node 0 to 8 instead of the > >> interleaving), I tried to measure the migration speed with and without > >> the patches on mainline. Its interesting.... > >> > >> 10000 pages | 100000 pages > >> -------------------------- > >> Mainline 165 ms 1674 ms > >> Mainline + first patch (move_pages) 191 ms 1952 ms > >> Mainline + all three patches 146 ms 1469 ms > >> > >> Though overall it gives performance improvement, some how it slows > >> down migration after the first patch. Will look into this further. > > > > What are you measuring actually? All pages migrated to the same node? > > The mount of time move_pages() system call took to move these many > pages from node 0 to node 8. Yeah they migrated to the same node. > > > Do you have any profiles? How stable are the results? > > No, are you referring to perf record kind profile ? Results were > repeating. Yes. I am really wondering because there souldn't anything specific to improve the situation with patch 2 and 3. Likewise the only overhead from the patch 1 I can see is the reduced batching of the mmap_sem. But then I am wondering what would compensate that later... -- Michal Hocko SUSE Labs -- 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