From owner-freebsd-net Sat Nov 10 12:52: 3 2001 Delivered-To: freebsd-net@freebsd.org Received: from inje.iskon.hr (inje.iskon.hr [213.191.128.16]) by hub.freebsd.org (Postfix) with ESMTP id EE15037B41A for ; Sat, 10 Nov 2001 12:51:59 -0800 (PST) Received: from tel.fer.hr (zg05-164.dialin.iskon.hr [213.191.138.165]) by mail.iskon.hr (8.11.4/8.11.4/Iskon 8.11.3-1) with ESMTP id fAAKpcu17578; Sat, 10 Nov 2001 21:51:39 +0100 (MET) Message-ID: <3BED9357.D4C48191@tel.fer.hr> Date: Sat, 10 Nov 2001 21:51:35 +0100 From: Marko Zec X-Mailer: Mozilla 4.78 [en] (Windows NT 5.0; U) X-Accept-Language: en MIME-Version: 1.0 To: Kyunghwan Kim Cc: freebsd-net@freebsd.org Subject: Re: Intel nics (fxp, wx, gx) questions References: <20011110141356.A60420@ada.snu.ac.kr> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-net@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org Kyunghwan Kim wrote: > Some questions related to intel nic drivers: > > 1. fxp microcode > Marko has summited fxp interrupt bundling patch before. > Is there any reference to write fxp microcode? No that I am aware of, as I just used the microcode supplied with Intel's Linux driver in binary form. After conducting a few TCP throughput tests with interrupt coalescing microcode, I found that depending of the specific environment, quite serious TCP perfomance degradation can follow as a result of additional delay introduced by this method. In my strong opinion, the proper way to go in reducing the interrupt overhead was directed by Luigi Rizzo's polling patch, as it does not suffer from TCP perf. degradation problems, yet still completely removes the interrupt overhead, and introduces some adittional goodies. Marko To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message