From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Paul E. McKenney" Subject: Re: [ANNOUNCE] 3.0.1-rt8 Date: Tue, 9 Aug 2011 18:08:37 -0700 Message-ID: <20110810010837.GC2737@linux.vnet.ibm.com> References: <1312580681.28695.44.camel@twins> <4E417D12.70608@localhost> <1312915646.1083.79.camel@twins> Reply-To: paulmck@linux.vnet.ibm.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Fernando Lopez-Lezcano , linux-kernel , Thomas Gleixner , linux-rt-users To: Peter Zijlstra Return-path: Received: from e7.ny.us.ibm.com ([32.97.182.137]:35812 "EHLO e7.ny.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750841Ab1HJBIj (ORCPT ); Tue, 9 Aug 2011 21:08:39 -0400 Content-Disposition: inline In-Reply-To: <1312915646.1083.79.camel@twins> Sender: linux-rt-users-owner@vger.kernel.org List-ID: On Tue, Aug 09, 2011 at 08:47:26PM +0200, Peter Zijlstra wrote: > On Tue, 2011-08-09 at 11:31 -0700, Fernando Lopez-Lezcano wrote: > > Hmmm, what are reasonable values for CONFIG_RCU_BOOST_PRIO and > > CONFIG_RCU_BOOST_DELAY? (or at least what are you using?). > > I'm using the defaults, since I haven't actually got a workload besides > compiling kernels ;-) > > Changing these values is very much workload dependent, the PRIO should > be high enough not to cause memory starvation, but low enough not to > disrupt anything important, and thus completely dependent on you > favourite RT workload. > > The same goes for the DELAY, too long and you run out of memory, too > short and you get more overhead, depends on your workload, your machine > memory size etc.. > > Paul should of course be put on trial for giving us these knobs, but > seeing where they come from I totally understand they exist ;-) You mean my experience at the tail end of 3.0 -wasn't- being put on trial? ;-) Thanx, Paul