From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from web33102.mail.mud.yahoo.com (web33102.mail.mud.yahoo.com [209.191.69.132]) by ozlabs.org (Postfix) with SMTP id E337ADDF54 for ; Sat, 21 Jun 2008 09:44:57 +1000 (EST) Date: Fri, 20 Jun 2008 16:44:55 -0700 (PDT) From: Sanjay Patel Subject: Re: [RFC 1/3] powerpc: __copy_tofrom_user tweaked for Cell To: benh@kernel.crashing.org In-Reply-To: <1214004027.8011.182.camel@pasglop> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Message-ID: <834245.69039.qm@web33102.mail.mud.yahoo.com> Cc: Mark Nelson , Gunnar von Boehn , Arnd Bergmann , linuxppc-dev@ozlabs.org, Michael Ellerman , cbe-oss-dev@ozlabs.org Reply-To: sanjay3000@yahoo.com List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , --- On Fri, 6/20/08, Benjamin Herrenschmidt wrote: > I though OS X had a trick with a CR bit that would disable > the dcbz optimization on the first alignment fault ? Or did they > totally remove it ? Ah, it's coming back to me. :) Apple added 'dcbz', removed it, and then there was the clever trick of optimizing the code path with a boot-time perf test and/or changing the code on the first fault...I'm not sure what's implemented in the recent builds. If Linux can do something similar, that should allow good perf on cacheable and cache-inhibited space as well as different CPUs (eg, if 'dcba' is available, then you don't need the alignment fault hack). --Sanjay