From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757528Ab2CSDrX (ORCPT ); Sun, 18 Mar 2012 23:47:23 -0400 Received: from mailout-de.gmx.net ([213.165.64.23]:52183 "HELO mailout-de.gmx.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1753150Ab2CSDrW (ORCPT ); Sun, 18 Mar 2012 23:47:22 -0400 X-Authenticated: #14349625 X-Provags-ID: V01U2FsdGVkX19IsFhux9ZIUGuawmaSycK3tefn9jrW1XCfGjeK1F emd5OAeaATXaPc Message-ID: <1332128832.7283.34.camel@marge.simpson.net> Subject: Re: [PATCH RFC] rcu: Permit limiting of force_quiescent_state() latency From: Mike Galbraith To: paulmck@linux.vnet.ibm.com Cc: linux-kernel@vger.kernel.org, mingo@elte.hu, laijs@cn.fujitsu.com, dipankar@in.ibm.com, akpm@linux-foundation.org, mathieu.desnoyers@efficios.com, josh@joshtriplett.org, niv@us.ibm.com, tglx@linutronix.de, peterz@infradead.org, rostedt@goodmis.org, Valdis.Kletnieks@vt.edu, dhowells@redhat.com, eric.dumazet@gmail.com, darren@dvhart.com, fweisbec@gmail.com, sivanich@sgi.com Date: Mon, 19 Mar 2012 04:47:12 +0100 In-Reply-To: <20120319001014.GD2393@linux.vnet.ibm.com> References: <20120316160328.GA7207@linux.vnet.ibm.com> <1332003451.8147.77.camel@marge.simpson.net> <1332082821.7291.27.camel@marge.simpson.net> <20120319001014.GD2393@linux.vnet.ibm.com> Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.2.1 Content-Transfer-Encoding: 7bit Mime-Version: 1.0 X-Y-GMX-Trusted: 0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sun, 2012-03-18 at 17:10 -0700, Paul E. McKenney wrote: > Thank you very much for your efforts on this!!! No problem, irq holdoff troubles are worth effort. > Given that you were seeing about 200-microsecond latency spikes on > grace-period initialization, I would expect that you would need about > 200 dyntick-idle CPUs for force_quiescent_state() to give you a > ten-microsecond spike, so I am not surprised that you could not see > the difference on 60 CPUs, which probably have given you something > like 3 microseconds.. Crawling over fewer locks should still save cycles, so I'll measure. Big box rt needs every little usec we can scrape together. -Mike