Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261681AbVCNSqX (ORCPT ); Mon, 14 Mar 2005 13:46:23 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261710AbVCNSpp (ORCPT ); Mon, 14 Mar 2005 13:45:45 -0500 Received: from e6.ny.us.ibm.com ([32.97.182.146]:39652 "EHLO e6.ny.us.ibm.com") by vger.kernel.org with ESMTP id S261681AbVCNSnu (ORCPT ); Mon, 14 Mar 2005 13:43:50 -0500 Date: Mon, 14 Mar 2005 10:42:59 -0800 From: "Martin J. Bligh" To: Arjan van de Ven , Andrew Morton cc: Christoph Lameter , linux-kernel@vger.kernel.org, linux-mm@vger.kernel.org, torvalds@osdl.org Subject: Re: [PATCH] Prefaulting Message-ID: <547140000.1110825779@flay> In-Reply-To: <1110618637.6292.36.camel@laptopd505.fenrus.org> References: <20050311172228.773cf03d.akpm@osdl.org> <1110618637.6292.36.camel@laptopd505.fenrus.org> X-Mailer: Mulberry/2.1.2 (Linux/x86) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1303 Lines: 30 --On Saturday, March 12, 2005 10:10:36 +0100 Arjan van de Ven wrote: > >> > From a quick peek it seems that the patch makes negligible difference for a >> kernel compilation when prefaulting 1-2 pages and slows the workload down >> quite a lot when prefaulting up to 16 pages. > > well the last time I saw prefaulting experiments (Ingo was involved > iirc) the problem was that the hitrate for the prefaults was such that > the costs for tearing down the extra redundant rmap chains was more > expensive than taking the "extra" faults. It seems linux has pretty > cheap faulting logic invalidating some of traditional OS assumptions... Yes - is pretty much exactly what I saw with this too, even if we only prefaulted up to 4 pages that were already in the pagecache. The additional cost in zap_pte_rage etc killed us, and it was wholly detrimental. I think we need better locality in glibc, etc before this is of much use. I had a debug patch somewhere to show how sparse the layout of ptes was, I'll see if I can find it. M. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/