From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753353Ab1FTPda (ORCPT ); Mon, 20 Jun 2011 11:33:30 -0400 Received: from smtp109.prem.mail.ac4.yahoo.com ([76.13.13.92]:40969 "HELO smtp109.prem.mail.ac4.yahoo.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1752956Ab1FTPdF (ORCPT ); Mon, 20 Jun 2011 11:33:05 -0400 X-Yahoo-Newman-Property: ymail-3 X-YMail-OSG: 20gQWUgVM1nJWhQlTsy7i3RjUODu63AwNLuYVX9zi3ebjvC Mn0HXaa_81IMN2a8PC9EzfoGRTwe0TCNLVwq9yH1sTabM5P5WENp60.S..Qp j2JEG0t2UXlhtRmm4U_UfdPTCXJtpfiP6Du1OGB9c5uNuewi.WJpUs1eQKoz SSo7ck0Qgf.47OpiPAd6c2n38eOKWrURmMBfzSuX8JInPf8ui7Y0KSS_Egy0 IOWSvxiBDwFGT_2Wn_aXAsSuJYIm8WZKYoa13M7hMXNYxSbCMuQWxkKIqqAL 4EcrDo40it.kzsHHOKljcCGyHjwW2HsM2bkQ6Lr1TQm2JyYFI X-Yahoo-SMTP: _Dag8S.swBC1p4FJKLCXbs8NQzyse1SYSgnAbY0- Message-Id: <20110620153244.214038140@linux.com> User-Agent: quilt/0.48-1 Date: Mon, 20 Jun 2011 10:32:44 -0500 From: Christoph Lameter To: Pekka Enberg Cc: David Rientjes Cc: Andi Kleen Cc: tj@kernel.org Cc: Metathronius Galabant Cc: Matt Mackall Cc: Eric Dumazet Cc: Adrian Drzewiecki Cc: linux-kernel@vger.kernel.org Subject: [slub p2 0/4] SLUB: [RFC] Per cpu partial lists V2 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org The following patchset applied on top of the lockless patchset V7. It introduces per cpu partial lists which allow a performance increase of around ~15 during contention for the nodelock (can be tested using hackbench). These lists help to avoid per nodelocking overhead. Allocator latency could be further reduced by making these operations work without disabling interrupts (like the fastpath and the free slowpath) as well as implementing better ways of handling ther cpu array with partial pages. I am still not satisfied with the cleanliness of the code after these changes. Some review with suggestions as to how to restructure the code given these changes in operations would be appreciated. It is interesting to note that BSD has gone to a scheme with partial pages only per cpu (source: Adrian). Transfer of cpu ownerships is done using IPIs. Probably too much overhead for our taste. The use of a few per cpu partial pages looks to be beneficial though. Note that there is no performance gain when there is no contention. Performance: Before After ./hackbench 100 process 200000 Time: 2299.072 1742.454 ./hackbench 100 process 20000 Time: 224.654 182.393 ./hackbench 100 process 20000 Time: 227.126 182.780 ./hackbench 100 process 20000 Time: 219.608 182.899 ./hackbench 10 process 20000 Time: 21.769 18.756 ./hackbench 10 process 20000 Time: 21.657 18.938 ./hackbench 10 process 20000 Time: 23.193 19.537 ./hackbench 1 process 20000 Time: 2.337 2.263 ./hackbench 1 process 20000 Time: 2.223 2.271 ./hackbench 1 process 20000 Time: 2.269 2.301