From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755515Ab0EaRwP (ORCPT ); Mon, 31 May 2010 13:52:15 -0400 Received: from mail-wy0-f174.google.com ([74.125.82.174]:36943 "EHLO mail-wy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754630Ab0EaRwN (ORCPT ); Mon, 31 May 2010 13:52:13 -0400 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=subject:from:to:cc:in-reply-to:references:content-type:date :message-id:mime-version:x-mailer:content-transfer-encoding; b=C5kdEuzB/NP6xp0vOpcTryWGaDVWyGSEnMSEThc94PMUNoLyvnTvt1nNzjXEiAOdPp 6v7ABZ8gdfMiAkOq/cJH68TIHLVmpHMB4gO96UWA7PlnUZTK3zjGZinq0vHcVYsMxoJi lG2QPwBlG7FXg75MMLO+7xWWSKS9qKhwQNm/0= Subject: Re: [PATCH] DM-CRYPT: Scale to multiple CPUs From: Eric Dumazet To: Andi Kleen Cc: herbert@gondor.hengli.com.au, dm-devel@redhat.com, linux-kernel@vger.kernel.org, agk@redhat.com, ak@linux.intel.com In-Reply-To: <20100531174611.GF10766@basil.fritz.box> References: <20100531160425.GA20344@basil.fritz.box> <1275323695.3291.122.camel@edumazet-laptop> <20100531174611.GF10766@basil.fritz.box> Content-Type: text/plain; charset="UTF-8" Date: Mon, 31 May 2010 19:52:08 +0200 Message-ID: <1275328328.3291.126.camel@edumazet-laptop> Mime-Version: 1.0 X-Mailer: Evolution 2.28.3 Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Le lundi 31 mai 2010 à 19:46 +0200, Andi Kleen a écrit : > On Mon, May 31, 2010 at 06:34:55PM +0200, Eric Dumazet wrote: > > Yes seems like I'm behind on per cpu innovations. > > > Since commit e0fdb0e050eae331, we have a __percpu annotation so that > > sparse can be augmented. This would also make the comment unnecessary... > > I think the comment is still necessary. > > > > > > + __get_cpu_var(io_wq_cpu) = current; > > > > this_cpu_write(io_wq_cpu, current) > > And this one looks more ugly than before and does not seem > to generate better code, so I decided to skip it. This is because you whould use __this_cpu_write() in this context of course ;) generated code : one instruction