home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Turnbull China Bikeride
/
Turnbull China Bikeride - Disc 2.iso
/
BARNET
/
ARMLINUX
/
MAIL
/
9807
/
000085_neil@causality.com _Fri Jul 3 00:20:30 1998.msg
< prev
next >
Wrap
Internet Message Format
|
1998-08-03
|
2KB
Return-Path: <neil@causality.com>
Received: from post.mail.demon.net (post-12.mail.demon.net [194.217.242.41])
by odie.barnet.ac.uk (8.8.6/8.8.6) with SMTP id AAA30146
for <willy@odie.barnet.ac.uk>; Fri, 3 Jul 1998 00:20:29 +0100
Received: from (causality.com) [158.152.208.192]
by post.mail.demon.net with esmtp (Exim 1.82 #2)
id 0yrsfF-0007UP-00; Thu, 2 Jul 1998 23:21:17 +0000
Sender: neil@odie.barnet.ac.uk
Message-ID: <359C1198.368A09AD@causality.com>
Date: Fri, 03 Jul 1998 00:02:48 +0100
From: "Neil A. Carson" <neil@causality.com>
Organization: Causality Limited
X-Mailer: Mozilla 4.04 [en] (X11; I; FreeBSD 3.0-CURRENT i386)
MIME-Version: 1.0
To: Matthew Wilcox <willy@odie.barnet.ac.uk>
CC: phil@oregan.net, linux-arm@vger.rutgers.edu
Subject: Re: ARM and RPM
References: <199807020947.KAA28766@odie.barnet.ac.uk>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Status: RO
Matthew Wilcox wrote:
> The Really Bad Thing is that on StrongARM, SWP does _not_ bypass the
> cache, which makes life harder for SMP systems. Of course, this isn't
> the only `interesting' feature of the StrongARM's cache, but this
> has been discussed ad nauseam previously on this list and those who
> are interested may grep the archives. One work-around is to put all
> semaphores in an uncached page. Another is to disable the data cache
> entirely and put a shared discrete cache on your SMP processor card,
> as Acorn are rumoured to be doing for Phoeb[l]e.
Good description of SWP, much better than mine! Semaphores have to go in
an uncached area with SA, I think. Bit silly really---but not
impossible.
Pretty unlikely I would think. As the bus clock can't run at more than
66MHz on current chips, and the current speed of main memory can exceed
this on bursts (100MHz), there's no point in putting a cache on the
outside---main memory may just as well be used instead :) If SA could
run its external bus at say 200MHz, then I guess one could implement an
external cache effectively. However, this is unlikely to happen. This
effectively means that efficient SMP with StrongARMs is impossible.
> If I've misheard a rumour, I'd like to apologise now and hope that
> someone will correct me authoritatively.
Not authoritative, just technical :)
Regards,
Neil
--
Neil A. Carson