From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Graf Subject: Re: [PATCH] improvement on net/sched/cls_fw.c's hash function Date: Tue, 5 Apr 2005 12:38:27 +0200 Message-ID: <20050405103827.GL26731@postel.suug.ch> References: <20050405133336.0247.LARK@linux.net.cn> <20050404223744.1f04c130.davem@davemloft.net> <20050405140342.024A.LARK@linux.net.cn> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: netdev@oss.sgi.com Return-path: To: Wang Jian Content-Disposition: inline In-Reply-To: <20050405140342.024A.LARK@linux.net.cn> Sender: netdev-bounce@oss.sgi.com Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org * Wang Jian <20050405140342.024A.LARK@linux.net.cn> 2005-04-05 14:05 > New patch attached. Hashsize is 256, the same as old one. Do you have any numbers that could prove that this change actually improves the hash distribution and thus the overall lookup performance? The most often used and thus most important range of mark values is definitely 0..255. I did not look into jhash but the risk of collisions definitely increases with this change which affects about 90% of the users of fw which could benefit from a collision free hashtable so far. I would appreciate if you could provide some numbers proving both the need and actual improvement of this change since fwmark is one of the most often used classifiers. Cheers