From owner-freebsd-hackers  Fri Dec 13 02:54:29 1996
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.4/8.8.4) id CAA08488
          for hackers-outgoing; Fri, 13 Dec 1996 02:54:29 -0800 (PST)
Received: from hda.hda.com (ip24-max1-fitch.ziplink.net [199.232.245.24])
          by freefall.freebsd.org (8.8.4/8.8.4) with SMTP id CAA08482
          for <Hackers@freebsd.org>; Fri, 13 Dec 1996 02:54:25 -0800 (PST)
Received: (from dufault@localhost) by hda.hda.com (8.6.12/8.6.12) id FAA12751; Fri, 13 Dec 1996 05:51:45 -0500
From: Peter Dufault <dufault@hda.com>
Message-Id: <199612131051.FAA12751@hda.hda.com>
Subject: Re: Implementing atomic DCAS on Intel for NBS
In-Reply-To: <199612130653.BAA17932@chai.plexuscom.com> from Bakul Shah at "Dec 13, 96 01:53:31 am"
To: bakul@plexuscom.com (Bakul Shah)
Date: Fri, 13 Dec 1996 05:51:44 -0500 (EST)
Cc: michaelh@cet.co.jp, Hackers@freebsd.org
X-Mailer: ELM [version 2.4ME+ PL25 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

> Comments about NBS:
> 
> In general, when the number of words read exceeds the number of
> written words (for shared data structures), NBS will be more
> efficient than lockIng.  Deadlock simply can not happen because
> atmost there is just one shared lock.

That one shared lock will get pretty busy, while with standard
locking the lock activity can "page" to a point in an MP hierarchy
where they are actually used.  I'll have to look up the references
to see how this locality issue is discussed.

Language support with some kind of commit operation on data structures
seems attractive - by this I mean only declaring those parts of a
data structure that must be lockable lockable, and automatically
getting write accesses nCAS'd where the system would update all
fields at once from a temporary (together with the serial number
update - go easy on this - I haven't thought about it much yet).

-- 
Peter Dufault (dufault@hda.com)   Real-Time Machine Control and Simulation
HD Associates, Inc.               Voice: 508 433 6936