From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner+w=401wt.eu-S1758623AbXEYITP@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1758623AbXEYITP (ORCPT <rfc822;w@1wt.eu>);
	Fri, 25 May 2007 04:19:15 -0400
Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752182AbXEYITG
	(ORCPT <rfc822;linux-kernel-outgoing>);
	Fri, 25 May 2007 04:19:06 -0400
Received: from holomorphy.com ([66.93.40.71]:35338 "EHLO holomorphy.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1751982AbXEYITE (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
	Fri, 25 May 2007 04:19:04 -0400
Date: Fri, 25 May 2007 01:19:44 -0700
From: William Lee Irwin III <wli@holomorphy.com>
To: Ingo Molnar <mingo@elte.hu>
Cc: Andi Kleen <andi@firstfloor.org>, Satyam Sharma <satyam.sharma@gmail.com>,
       Andrew Morton <akpm@linux-foundation.org>, linux-kernel@vger.kernel.org,
       Peter Zijlstra <a.p.zijlstra@chello.nl>
Subject: Re: [patch] i386, numaq: enable TSCs again
Message-ID: <20070525081944.GZ31925@holomorphy.com>
References: <20070525071005.GA6431@elte.hu> <a781481a0705250022t3dec8f13u7c75d0ade40b0010@mail.gmail.com> <20070525073158.GB8094@one.firstfloor.org> <20070525073957.GA15207@elte.hu> <20070525074316.GA16821@elte.hu> <20070525074915.GA18400@elte.hu> <20070525075446.GA20140@elte.hu> <20070525080827.GA22399@elte.hu>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20070525080827.GA22399@elte.hu>
Organization: The Domain of Holomorphy
User-Agent: Mutt/1.5.13 (2006-08-11)
Sender: linux-kernel-owner@vger.kernel.org
X-Mailing-List: linux-kernel@vger.kernel.org

On Fri, May 25, 2007 at 10:08:27AM +0200, Ingo Molnar wrote:
> Andi, Andrew, do you remember why we disabled TSCs on NUMAQ? It was 
> slightly async between CPUs, right? In that case we should try the patch 
> below.

I remember. It was far beyond "slightly async;" they would drift
minutes apart during reasonable amounts of uptime, though it would take
at least several days to drift so far (I don't recall how long it took).

TSC synchronization is uniformly impossible on NUMA-Q. Bootlogs showing
the results of the attempts are still extant. They shouldn't end up too
far apart right after booting, but I don't have even ballpark estimates.
I'd hazard a guess of a few seconds.

NUMA-Q's also supported mixed CPU models in hardware, though that's
not really expected to be handled by Linux. I suspect DYNIX/ptx would
be used by anyone interested in that.


-- wli