[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Wait-free Synchronization, RTLinux and CSP

Thanks for the information, Adrian... It will be a great 
disappointment if the RTLinux-like approach is killed by an
"openness" dispute. I hadn't heard of that... And I did not get
time to examine the technical specs closely, so the "nonblocking"
notion comes as a surprise to me and doesn't sound very workable.
I hope other people on the list dig into this because I never
have time. The ideal would be to get it working like high
priority code on a Transputer - and to introduce a small, DOS-like
OS that would allow easy development of "all-powerful" code without
the awful constraints of "device driver specs". I have always believed
that we could occupy a very VALUABLE niche between "all-powerful DOS"
programs with their lack of robustness, and "device drivers" with their
extremely restrictive, complex specs which make them useless for any
quick, clean effort.


>From a.e.lawrence@xxxxxxxxxxx Wed Mar  6 03:27:31 2002
Date: Wed, 06 Mar 2002 11:27:34 +0000
From: A E Lawrence <A.E.Lawrence@xxxxxxxxxxx>
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv: Gecko/20010901
X-Accept-Language: en-us
MIME-Version: 1.0
To: Lawrence Dickson <tjoccam@xxxxxxxxxxxxx>
CC: occam-com@xxxxxxxxx
Subject: Re: Wait-free Synchronization, RTLinux and CSP
References: <200203051829.g25IT5c74107@xxxxxxxxxxxxx>
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit
X-Lboro-Filtered: mailer2.lut.ac.uk, Wed, 6 Mar 2002 11:27:27 +0000

Lawrence Dickson wrote:

> I did mention RTLinux a few months (?) back. It is actually not Linux,
> but an all-powerful and very small embedded environment over which 
> Linux runs as the lowest priority process. This is done with a clever
> adaptation of the Linux macros CLI and STI. Linux communicates with
> embedded "masters" through pipes which behave much like channels. I 
> think it is a natural both for development of a super-DOS and of a
> CSP-based superstructure including occam - with complete mastery of
> the machine at the same time as all the tools of Linux are available.

It was your post that I remembered, I think.

RTLinux seems to be out of favour with the majority of the Linux 
community over a patent issue and openness.

Most of the current work is on RTAI


which uses a similar approach, indeed forked off RTLinux.

At first sight, RTLinux is not obviously CSP friendly in that its 
communication model is non-blocking and based on shared memory objects.
But that is associated with  wait-free synchronisation which I think is 
  inherently buffered, at least for the "universal" objects. I don't 
think there is any problem in describing them as suitable CSP processes, 
but I
have been trying to understand the various papers where the ideas are 
introduced. They seem to lack simple examples which does not help.

The only books that seem to mention the ideas are Attiya & Welch and Lynch.