From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Zijlstra Date: Tue, 06 Jan 2009 23:16:03 +0000 Subject: Re: [PATCH] configure HAVE_UNSTABLE_SCHED_CLOCK for SGI_SN systems Message-Id: <1231283763.11687.135.camel@twins> List-Id: References: <20090106162741.GA7991@sgi.com> <57C9024A16AD2D4C97DC78E552063EA35CB95575@orsmsx505.amr.corp.intel.com> <20090106201950.GA3850@sgi.com> <57C9024A16AD2D4C97DC78E552063EA35CB955B4@orsmsx505.amr.corp.intel.com> <1231275441.11687.110.camel@twins> <20090106225054.GB3850@sgi.com> In-Reply-To: <20090106225054.GB3850@sgi.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Robin Holt Cc: "Luck, Tony" , Dimitri Sivanich , "linux-ia64@vger.kernel.org" , Greg KH , "linux-kernel@vger.kernel.org" , Gregory Haskins , Nick Piggin , Tony Luck On Tue, 2009-01-06 at 16:50 -0600, Robin Holt wrote: > On Tue, Jan 06, 2009 at 09:57:20PM +0100, Peter Zijlstra wrote: > > On Tue, 2009-01-06 at 12:34 -0800, Luck, Tony wrote: > > > > > All ia64 systems are potentially affected ... but perhaps you might > > > > > never see the problem on most because the itc clocks are synced as close > > > > > as s/w can get them when cpus are brought on line. > > > > > > > > Do you want Dimitri to resubmit with this set for all IA64 or leave it > > > > as is? > > > > > > I'd like to understand the impact of turning on HAVE_UNSTABLE_SCHED_CLOCK > > > > > > It looks like both the i386_defconfig and x86_64_defconfig choose this, > > > so at least ia64 will be hitting the well tested code paths > > > > > > Have the other architectures just not hit this yet? Or do they all have > > > "stable" sched_clock() functions? > > > > > > > > > sched_clock() seemed like such a straightforward thing to begin with. A > > > quick & easy way to measure a time delta ON THE SAME CPU. I'm not at > > > all sure why it has been co-opted for general time measurement. > > > > It came from the complication of needing to tell a remote cpu's time due > > to remote wakeups in the scheduler. > > But doesn't scheduler tick advance the rq->clock? Why do the others > need to fiddle with a remote runqueue's clock? When that cpu starts > taking ticks again, it will update it's rq->clock field and start the > processes. I guess I am a lot underinformed about the new scheduler > design. We try to do better than tick based time accounting these days.