From owner-freebsd-current Sun Nov 10 15:41:29 1996 Return-Path: owner-current Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id PAA00547 for current-outgoing; Sun, 10 Nov 1996 15:41:29 -0800 (PST) Received: from who.cdrom.com (who.cdrom.com [204.216.27.3]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id PAA00537; Sun, 10 Nov 1996 15:41:27 -0800 (PST) Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.211]) by who.cdrom.com (8.7.5/8.6.11) with SMTP id OAA26827 ; Sun, 10 Nov 1996 14:50:07 -0800 (PST) Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id PAA17065; Sun, 10 Nov 1996 15:40:42 -0700 From: Terry Lambert Message-Id: <199611102240.PAA17065@phaeton.artisoft.com> Subject: Re: Bad news: kind-of about page coloring To: dyson@freebsd.org Date: Sun, 10 Nov 1996 15:40:42 -0700 (MST) Cc: current@freebsd.org In-Reply-To: <199611101926.OAA06336@dyson.iquest.net> from "John S. Dyson" at Nov 10, 96 02:26:05 pm X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk > As some of you know, I just got a PP motherboard (mostly to be able > to work on SMP in the near future.) For now, I just have one processor, > and have been studying it (in relation to FreeBSD.) As I kind-of expected > due to the 4 way 2nd level cache on the PP, the page coloring code appears > to be more of a hinderance than an advantage. I did find a significant > improvement on the Pentium though. Just to let you know, I am going to > make the coloring code work by setting itself up at runtime as opposed > to compile time. The coloring code does have a liability of messing up > our page caching stats (a little bit), but since the the PP cache is > 4 way associative, we don't need the coloring nearly as much. So, IMO, > coloring appears to be generally bad on the PPro. I'm a bit curious: how would you gage the associativity of a two processor system? Ie: is it possible that an N processor (N>1) PP system would still benefit from coloring if it was layered on top of the coherency model? I assume the code is on top of the coherency model, in any case... Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.