Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263044AbUCMEzH (ORCPT ); Fri, 12 Mar 2004 23:55:07 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263046AbUCMEzH (ORCPT ); Fri, 12 Mar 2004 23:55:07 -0500 Received: from gate.crashing.org ([63.228.1.57]:20441 "EHLO gate.crashing.org") by vger.kernel.org with ESMTP id S263044AbUCMEzD (ORCPT ); Fri, 12 Mar 2004 23:55:03 -0500 Subject: Re: [PATCH] ppc32 copy_to_user dcbt fixup From: Benjamin Herrenschmidt To: Bryan Rittmeyer Cc: Linux Kernel list , linuxppc-dev list , Paul Mackerras In-Reply-To: <20040313041547.GB11512@staidm.org> References: <20040313041547.GB11512@staidm.org> Content-Type: text/plain Message-Id: <1079153403.2348.82.camel@gaston> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.5 Date: Sat, 13 Mar 2004 15:50:03 +1100 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 886 Lines: 22 > [1] csum_partial_copy_generic does not use dcbt/dcbz despite being > scorching hot in TCP workloads. I'm cooking up another patch to > dcb?ize it. > > [2] http://staidm.org/linux/ppc/copy_dcbt/copyuser-microbench.tar.bz2 I'll have a good look at it asap. Also, keep in mind that between the dcbz and the time you can actually write to that cache line, some CPUs may need some time to complete the L1 dcbz operation, which can lead to poor performances, at least I've been told this is the case on POWER3, maybe others. It would be wise to make the dcbz as long as possible in "advance" before the actual write to the cache line. Ben. - 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/