From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andi Kleen Subject: Re: [dm-devel] [PATCH] DM-CRYPT: Scale to multiple CPUs Date: Wed, 02 Jun 2010 09:51:59 +0200 Message-ID: <4C060D9F.3050600@linux.intel.com> References: <20100531160425.GA20344@basil.fritz.box> <20100601043901.GA25693@gondor.apana.org.au> <20100602051403.GA5784@gondor.apana.org.au> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: Mikulas Patocka Cc: Herbert Xu , device-mapper development , linux-kernel@vger.kernel.org, agk@redhat.com List-Id: dm-devel.ids >> >>> It is pointless to track the submitting CPU. >> >> No you are wrong. > > For what? For avoiding cache bounces? But the encrypting is > order-of-magnitude slower than memory speed. On a system with reasonably fast CPUs it's fastest to just stay on your current CPU, don't try to talk to other CPUs, avoid communication, just get the work done ASAP. But make sure you can do this on multiple CPUs at the same time. With AES-NI this becomes even more pronounced because it effectively makes the CPU faster for encryption. -andi