Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S262324AbVCPKaW (ORCPT ); Wed, 16 Mar 2005 05:30:22 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262341AbVCPKaW (ORCPT ); Wed, 16 Mar 2005 05:30:22 -0500 Received: from mx1.elte.hu ([157.181.1.137]:58287 "EHLO mx1.elte.hu") by vger.kernel.org with ESMTP id S262324AbVCPKaG (ORCPT ); Wed, 16 Mar 2005 05:30:06 -0500 Date: Wed, 16 Mar 2005 11:29:51 +0100 From: Ingo Molnar To: Andrew Morton Cc: rostedt@goodmis.org, rlrevell@joe-job.com, linux-kernel@vger.kernel.org Subject: Re: [patch 0/3] j_state_lock, j_list_lock, remove-bitlocks Message-ID: <20050316102951.GA18247@elte.hu> References: <20050315120053.GA4686@elte.hu> <20050315133540.GB4686@elte.hu> <20050316085029.GA11414@elte.hu> <20050316011510.2a3bdfdb.akpm@osdl.org> <20050316095155.GA15080@elte.hu> <20050316020408.434cc620.akpm@osdl.org> <20050316101209.GA16893@elte.hu> <20050316022638.237f72cd.akpm@osdl.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050316022638.237f72cd.akpm@osdl.org> User-Agent: Mutt/1.4.1i X-ELTE-SpamVersion: MailScanner 4.31.6-itk1 (ELTE 1.2) SpamAssassin 2.63 ClamAV 0.73 X-ELTE-VirusStatus: clean X-ELTE-SpamCheck: no X-ELTE-SpamCheck-Details: score=-4.9, required 5.9, autolearn=not spam, BAYES_00 -4.90 X-ELTE-SpamLevel: X-ELTE-SpamScore: -4 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1042 Lines: 27 * Andrew Morton wrote: > I forget how much of the 1000% came from that, but it was quite a lot. > > Removing the BKL was the first step. That took the context switch > rate under high load from ~10,000/sec up to ~300,000/sec. Because the > first thing a CPU hit on entry to the fs was then a semaphore. > Performance rather took a dive. > > Of course the locks also became much finer-grained, so the contention > opportunities lessened. But j_list_lock and j_state_lock have fs-wide > scope, so I'd expect the context switch rate to go up quite a lot > again. > > The hold times are short, and a context switch hurts rather ore than a > quick spin. which particular workload was this - dbench? (I can try PREEMPT_RT on an 8-way, such effects will show up tenfold.) Ingo - 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/