From mboxrd@z Thu Jan 1 00:00:00 1970 From: Changli Gao Subject: Re: GRO after RPS? Date: Mon, 26 Apr 2010 09:40:35 +0800 Message-ID: References: <20100425.170933.190068177.davem@davemloft.net> <20100426004934.GA12525@gondor.apana.org.au> <20100425.181757.10149105.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: herbert@gondor.apana.org.au, netdev@vger.kernel.org To: David Miller Return-path: Received: from mail-iw0-f178.google.com ([209.85.223.178]:65533 "EHLO mail-iw0-f178.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751796Ab0DZBk4 convert rfc822-to-8bit (ORCPT ); Sun, 25 Apr 2010 21:40:56 -0400 Received: by iwn8 with SMTP id 8so4526567iwn.16 for ; Sun, 25 Apr 2010 18:40:55 -0700 (PDT) In-Reply-To: <20100425.181757.10149105.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-ID: On Mon, Apr 26, 2010 at 9:17 AM, David Miller wro= te: > > The goal is to eliminate all packet header references from the pre-RP= S > path, and let the post-RPS cpu do it. If the NIC doesn't provide rxhash, RPS will have to compute one by one by itself. Is the hash computation more expensive than GRO? I think the hash computation is cheaper than GRO, so we can do RPS ASAP to avoid the direct CPU overload. --=20 Regards=EF=BC=8C Changli Gao(xiaosuo@gmail.com)