Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1750918AbWATM0Q (ORCPT ); Fri, 20 Jan 2006 07:26:16 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750919AbWATM0Q (ORCPT ); Fri, 20 Jan 2006 07:26:16 -0500 Received: from holly.csn.ul.ie ([136.201.105.4]:44935 "EHLO holly.csn.ul.ie") by vger.kernel.org with ESMTP id S1750912AbWATM0P (ORCPT ); Fri, 20 Jan 2006 07:26:15 -0500 Date: Fri, 20 Jan 2006 12:25:04 +0000 (GMT) From: Mel Gorman X-X-Sender: mel@skynet To: Yasunori Goto Cc: KAMEZAWA Hiroyuki , Joel Schopp , linux-mm@kvack.org, linux-kernel@vger.kernel.org, lhms-devel@lists.sourceforge.net Subject: Re: [Lhms-devel] Re: [PATCH 0/5] Reducing fragmentation using zones In-Reply-To: <20060120210353.1269.Y-GOTO@jp.fujitsu.com> Message-ID: References: <43D03C24.5080409@jp.fujitsu.com> <20060120210353.1269.Y-GOTO@jp.fujitsu.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2445 Lines: 61 On Fri, 20 Jan 2006, Yasunori Goto wrote: > > What sort of tests would you suggest? The tests I have been running to > > date are > > > > "kbuild + aim9" for regression testing > > > > "updatedb + 7 -j1 kernel compiles + highorder allocation" for seeing how > > easy it was to reclaim contiguous blocks > > BTW, is "highorder allocation test" your original test code? > If so, just my curious, I would like to see it too. ;-). > 1. Download http://www.csn.ul.ie/~mel/projects/vmregress/vmregress-0.20.tar.gz 2. Extract it to /usr/src/vmregress (i.e. there should be a /usr/src/vmregress/bin directory) 3. Download linux-2.6.11.tar.gz to /usr/src 4. Make a directory /usr/src/bench-stresshighalloc-test 5. cd to /usr/src/vmregress and run 3. cd to the directory and run ./configure --with-linux=/path/to/running/kernel make 5. Run the test bench-stresshighalloc.sh -z -k 6 --oprofile -z Will test using high memory -k 6 will build 1 kernel + 6 additional ones By default, it will try and allocate 275 order-10 pages. Specify the number of pages with -c and the order with -s The paths above are default paths. They can all be overridden with command line parameters like -t to specify a different kernel to use and -b to specify a different path to build all the kernels in. By default, the results will be logged to a directory whose name is based on the kernel being tested. For example, one result directory is ~/vmregressbench-2.6.16-rc1-mm1-clean/highalloc-heavy/log.txt Comparisions between different runs can be analysed by using diff-highalloc.sh. e.g. diff-highalloc.sh vmregressbench-2.6.16-rc1-mm1-clean vmregressbench-2.6.16-rc1-mm1-mbuddy-v22 If you want to test just high-order allocations while some other workload is running, use bench-plainhighalloc.sh. See --help for a list of available options. If you want to use bench-aim9.sh, download and build aim9 in /usr/src/aim9 and edit the s9workfile to specify the tests you are interested in. Use diff-aim9.sh to compare different runs of aim9. -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab - 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/