From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Tue, 15 Aug 2000 14:28:33 +0200 (CEST) From: Mike Galbraith Subject: Re: [prePATCH] new VM for linux-2.4.0-test4 In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-linux-mm@kvack.org Return-Path: To: Rik van Riel Cc: linux-mm@kvack.org, linux-kernel@vger.rutgers.edu List-ID: On Mon, 14 Aug 2000, Rik van Riel wrote: > OK, I overlooked one of the bad bad bad mistakes watashi > saw .. here is an -incremental- patch to fix the last > possible source of memory leakage... Hi Rik, I put vm6 and this bugfix into test7-pre4.ikd, checked for leakage with memleak.. found absolutely nothing. I then disabled ikd and did some light performance comparison using my favorite generic test (make -j30 bzImage [1]). Tests conducted in identical as possible manner compiling the same tree. test7-pre4.vm6 real 9m42.191s user 6m31.440s sys 0m34.820s test7-pre4.stock real 13m38.449s user 6m31.000s sys 0m38.250s ac22-classzone+ real 7m48.594s user 6m30.750s sys 0m31.860s Definite improvement over stock vm, but still not as good at keeping 30 hungry tasks fed as classzone (on my 128mb single PIII box). All numbers fully repeatable +- normal test jitter. Streaming I/O seems to be suffering a bit, but I didn't measure enough to be 100% sure of that. -Mike 1. Think of it as a simulation of thirty students on a small classroom server in a 3rd world nation (ala Brooklyn N.Y.;) all compiling their individual bits of a group project. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux.eu.org/Linux-MM/