Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S269103AbUINB77 (ORCPT ); Mon, 13 Sep 2004 21:59:59 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S269104AbUINB77 (ORCPT ); Mon, 13 Sep 2004 21:59:59 -0400 Received: from smtp200.mail.sc5.yahoo.com ([216.136.130.125]:2976 "HELO smtp200.mail.sc5.yahoo.com") by vger.kernel.org with SMTP id S269103AbUINB75 (ORCPT ); Mon, 13 Sep 2004 21:59:57 -0400 Message-ID: <4146508F.9080700@yahoo.com.au> Date: Tue, 14 Sep 2004 11:59:43 +1000 From: Nick Piggin User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.2) Gecko/20040820 Debian/1.7.2-4 X-Accept-Language: en MIME-Version: 1.0 To: Marcelo Tosatti CC: William Lee Irwin III , Andrew Morton , linux-kernel@vger.kernel.org Subject: Re: [pagevec] resize pagevec to O(lg(NR_CPUS)) References: <20040909163929.GA4484@logos.cnet> <20040909155226.714dc704.akpm@osdl.org> <20040909230905.GO3106@holomorphy.com> <20040909162245.606403d3.akpm@osdl.org> <20040910000717.GR3106@holomorphy.com> <414133EB.8020802@yahoo.com.au> <20040910174915.GA4750@logos.cnet> <20040912045636.GA2660@holomorphy.com> <4143D07E.3030408@yahoo.com.au> <20040913222127.GA23588@logos.cnet> In-Reply-To: <20040913222127.GA23588@logos.cnet> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1068 Lines: 31 Marcelo Tosatti wrote: >>Sure. And when you fill it with pages, they'll use up 32KB of dcache >>by using a single 64B line per page. Now that you've blown the cache, >>when you go to move those pages to another list, you'll have to pull >>them out of L2 again one at a time. >> >>OK, so a 511 item pagevec is pretty unlikely. How about a 64 item one >>with 128 byte cachelines, and you're touching two cachelines per >>page operation? That's 16K. > > > Nick, > > Note that you dont read/write data to the actual pages most of the > times pagevec's are used. The great majority is just page management code. > So we dont really blow the caches like you said. > You're often pulling them off lists though which is what will do it. Not the actual page, but the struct page. > I agree we need more tests :) > Yep. - 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/