Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!usc!apple!bbn.com!dcousins
From: dcousins@bbn.com (Dave Cousins)
Newsgroups: comp.realtime
Subject: Re: Spin locks
Message-ID: <62396@bbn.BBN.COM>
Date: 29 Jan 91 19:29:17 GMT
References: <830@marvin.jpl.oz> <1991Jan22.084616.16547@hemel.bull.co.uk> <1991Jan22.133107.934@msuinfo.cl.msu.edu>
Sender: news@bbn.com
Lines: 13

fox@rudolf.nscl.msu.edu (Ron Fox) writes:

>--
>operation, and you need some external thing (either an event, included
>here is a pre-empting process), or another CPU to break the loop.

I use spin locks on a shared memory architecture, basically because
the interrupt processing on our system was not fast enough to use as
a semaphore. Spin locks have low latency, especially when the system
releasing the lock (i.e. writing the lock bit) does so by cycle-stealling
from the polling cpu.

Dave Cousins