Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S269438AbUICA3L (ORCPT ); Thu, 2 Sep 2004 20:29:11 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S269409AbUICA27 (ORCPT ); Thu, 2 Sep 2004 20:28:59 -0400 Received: from viper.oldcity.dca.net ([216.158.38.4]:2709 "HELO viper.oldcity.dca.net") by vger.kernel.org with SMTP id S269438AbUICAYr (ORCPT ); Thu, 2 Sep 2004 20:24:47 -0400 Subject: Re: [patch] voluntary-preempt-2.6.9-rc1-bk4-R0 From: Lee Revell To: Ingo Molnar Cc: linux-kernel , "K.R. Foley" , Felipe Alfaro Solana , Daniel Schmitt , Mark_H_Johnson@raytheon.com, "P.O. Gaillard" In-Reply-To: <20040902221402.GA29434@elte.hu> References: <20040902063335.GA17657@elte.hu> <20040902065549.GA18860@elte.hu> <20040902111003.GA4256@elte.hu> <20040902215728.GA28571@elte.hu> <1094162812.1347.54.camel@krustophenia.net> <20040902221402.GA29434@elte.hu> Content-Type: text/plain Message-Id: <1094171082.19760.7.camel@krustophenia.net> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Thu, 02 Sep 2004 20:24:43 -0400 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1526 Lines: 36 On Thu, 2004-09-02 at 18:14, Ingo Molnar wrote: > * Lee Revell wrote: > > > > i've given up on the netdev_backlog_granularity approach, and as a > > > replacement i've modified specific network drivers to return at a safe > > > point if softirq preemption is requested. > > > > Makes sense, netdev_max_backlog never made a difference on my system > > (via-rhine driver). > > via-rhine does RX processing from the hardirq handler, this codepath is > harder to break up. The NAPI ->poll functions used by e100 and 8193too > are much easier to break up because RX throttling and re-trying is a > basic property of NAPI. > Judging from these graphs, all of the latency issues are solved, at least on my UP hardware, and the latencies seem to be getting very close to the limits of what the hardware can do: http://krustophenia.net/testresults.php?dataset=2.6.9-rc1-Q6#/var/www/2.6.9-rc1-Q6/jack-test-1 The worst case latency is only 160 usecs, and the vast majority fall into the pattern from 0 to 120 usecs. All of the spikes above 120 are almost certainly caused by netdev_max_backlog. However these are not long enough to cause any problems with my workload; the lowest practical latency for audio work is around 0.66 ms (32 frames at 48khz). Lee - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/