From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1759148AbYARFyG (ORCPT ); Fri, 18 Jan 2008 00:54:06 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752902AbYARFxz (ORCPT ); Fri, 18 Jan 2008 00:53:55 -0500 Received: from mx1.suse.de ([195.135.220.2]:34148 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752706AbYARFxy (ORCPT ); Fri, 18 Jan 2008 00:53:54 -0500 To: Kyle McMartin Cc: hpa@zytor.com, tglx@linutronix.de, mingo@redhat.com, linux-kernel@vger.kernel.org Subject: Re: [PATCH] x86: make clflush a required feature on x86_64 From: Andi Kleen References: <20080118045921.GB974@phobos.i.cabal.ca> Date: Fri, 18 Jan 2008 06:53:53 +0100 In-Reply-To: <20080118045921.GB974@phobos.i.cabal.ca> (Kyle McMartin's message of "Thu\, 17 Jan 2008 23\:59\:21 -0500") Message-ID: User-Agent: Gnus/5.11 (Gnus v5.11) Emacs/22.1 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Kyle McMartin writes: > Hopefully nobody will be stupid enough to implement a cpu without > it. Frankly, it seems safe enough given we already require SSE2. > > This means the compiler can optimise away "if (!cpu_has_clflush)" > blocks. The original required CPUID bit set for x86-64 came out of very detailed discussions and thinking from AMD. I think they had good reasons to chose the current set. I would only recommend to depart from it for very good reasons and frankly you haven't given any. You're talking about removing two instructions. One problem that we had in the past is that some simulators only implement the absolutely minimum feature set and you might have well broken one of these with this. -Andi