From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264791AbUGMKaM (ORCPT ); Tue, 13 Jul 2004 06:30:12 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264781AbUGMKaM (ORCPT ); Tue, 13 Jul 2004 06:30:12 -0400 Received: from mx1.redhat.com ([66.187.233.31]:24522 "EHLO mx1.redhat.com") by vger.kernel.org with ESMTP id S264795AbUGMK3Z (ORCPT ); Tue, 13 Jul 2004 06:29:25 -0400 From: David Howells In-Reply-To: <20040712175605.GA1735@rx8.austin.ibm.com> References: <20040712175605.GA1735@rx8.austin.ibm.com> To: "Jose R. Santos" Cc: akpm@osdl.org, linux-kernel@vger.kernel.org, slprat@us.ibm.com Subject: Re: [PATCH] Making i/dhash_entries cmdline work as it use to. User-Agent: EMH/1.14.1 SEMI/1.14.5 (Awara-Onsen) FLIM/1.14.5 (Demachiyanagi) APEL/10.6 Emacs/21.3 (i386-redhat-linux-gnu) MULE/5.0 (SAKAKI) MIME-Version: 1.0 (generated by SEMI 1.14.5 - "Awara-Onsen") Content-Type: text/plain; charset=US-ASCII Date: Tue, 13 Jul 2004 11:29:13 +0100 Message-ID: <2443.1089714553@redhat.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Jose R. Santos wrote: > Also, any particular reason why MAX_SYS_HASH_TABLE_ORDER was set to 14? > I am already seeing the need to go higher on my 64GB setup and was > wondering if this could be bumped up to 19. Yes. IBM did some testing and found that was about optimal. No significant gain was found with anything greater. > I'm sending a patch that get the cmdline options working as the did before > where the could override the kernel calculations and increases > MAX_SYS_HASH_TABLE_ORDER to 19. Only tested on PPC64 at the moment. You need to be careful increasing the maximum order - you have to remember that this affects several tables (well, at least two at the moment), and so the effect is multiplied. It may be reasonable to let the kernel cmdline override the maximum number of buckets calculated on the scaling factor provided to the function (effectively number of buckets per unit memory), but consider that the number of objects that can be allocated and linked into the table is in effect governed by such a factor. David