From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261921AbUCDO5r (ORCPT ); Thu, 4 Mar 2004 09:57:47 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261924AbUCDO5r (ORCPT ); Thu, 4 Mar 2004 09:57:47 -0500 Received: from atrey.karlin.mff.cuni.cz ([195.113.31.123]:31465 "EHLO atrey.karlin.mff.cuni.cz") by vger.kernel.org with ESMTP id S261921AbUCDO5o (ORCPT ); Thu, 4 Mar 2004 09:57:44 -0500 Date: Wed, 3 Mar 2004 18:34:11 +0100 From: Pavel Machek To: Timothy Miller Cc: "Nakajima, Jun" , richard.brunner@amd.com, linux-kernel@vger.kernel.org Subject: Re: Intel vs AMD64 Message-ID: <20040303173411.GC531@openzaurus.ucw.cz> References: <7F740D512C7C1046AB53446D37200173EA28A5@scsmsx402.sc.intel.com> <403E4681.20603@techsource.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <403E4681.20603@techsource.com> User-Agent: Mutt/1.3.27i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Hi! > >Yes, "implementation specific" is one of the differences between > >IA-32e > >and AMD64, i.e. that behavior is architecturally defined on AMD64, > >but > >on IA-32e (as I posted): > > Near branch with 66H prefix: > > As documented in PRM the behavior is implementation specific and > >should > > avoid using 66H prefix on near branches. > > > In other words, Intel's implementation deviates from the architecture > as defined by AMD. So it's not 100% compatible. I just want this > point to be clear. > > > If these sorts of branches are common enough (and I suspect they > are), then this sort of deviation could have a notable code-size (and They are not. -- 64 bytes from 195.113.31.123: icmp_seq=28 ttl=51 time=448769.1 ms