Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S262129AbVBCXpF (ORCPT ); Thu, 3 Feb 2005 18:45:05 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261673AbVBCXiv (ORCPT ); Thu, 3 Feb 2005 18:38:51 -0500 Received: from ozlabs.org ([203.10.76.45]:11230 "EHLO ozlabs.org") by vger.kernel.org with ESMTP id S263528AbVBCXid (ORCPT ); Thu, 3 Feb 2005 18:38:33 -0500 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <16898.46622.108835.631425@cargo.ozlabs.ibm.com> Date: Fri, 4 Feb 2005 10:39:10 +1100 From: Paul Mackerras To: Rik van Riel Cc: Marcelo Tosatti , Christoph Lameter , David Woodhouse , linux-mm@kvack.org, linux-kernel@vger.kernel.org, akpm@osdl.org Subject: Re: A scrub daemon (prezeroing) In-Reply-To: References: <1106828124.19262.45.camel@hades.cambridge.redhat.com> <20050202153256.GA19615@logos.cnet> <20050202163110.GB23132@logos.cnet> X-Mailer: VM 7.19 under Emacs 21.3.1 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 810 Lines: 22 Rik van Riel writes: > I'm not convinced. Zeroing a page takes 2000-4000 CPU > cycles, while faulting the page from RAM into cache takes > 200-400 CPU cycles per cache line, or 6000-12000 CPU > cycles. On my G5 it takes ~200 cycles to zero a whole page. In other words it takes about the same time to zero a page as to bring in a single cache line from memory. (PPC has an instruction to establish a whole cache line of zeroes in modified state without reading anything from memory.) Thus I can't see how prezeroing can ever be a win on ppc64. Regards, Paul. - 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/