Newsgroups: comp.sys.transputer
From: k3mi@jupiter.sun.csd.unb.ca (Brian d'Auriol)
Subject: process activation (email correspondance)
Organization: University of New Brunswick, Fredericton, NB, Canada
Date: Thu, 10 Mar 1994 17:48:54 GMT
Message-ID: <1994Mar10.174854.14131@jupiter.sun.csd.unb.ca>

************************************

The address below is not valid, i am forwarding this to the newsgroup.
brian d'Auriol

On Thu, 10 Mar 1994 ABack@dcs.exeter.ac.uk wrote:

> I'm not too sure about OCCAM syntax, but I know about T8 assembly and
> programming using the Imos C toolset.
> 
> As to what happens with the hardware process queues, when a process
> blocks waiting for communication, its workspace pointer does not
> appear in the process queue.  Rather its workspace pointer is stored
> in the channel (its just a word).  When the process doing the other
> half of the communication tries to read/ write from/to the channel it
> looks at the channel word, and if there is a process waiting it does
> the communication, and then puts the processes workspace pointer on to
> the end of the hardware process queue of appropriate level.
> 
> I'm not sure I answered your question, if I have missed the point tell
> me.
> 
> Adam
> 
> 
thanks for the response. i am quite aware of the concurrent requirements
of occam code.  my question relates to what happens when two processes
syncronize on the communication.  your answer was that the wptr is
put on the end of the queue.  i agree, this is my understanding as
well.  unfortunately, my timing results seem to indicate otherwise.
I am trying to ascertain if the 'transputer' behaves as is 
expected or if it has another mmmmm..... 'personality'.

thanks for the response


------------------------------ End of forwarded message 1


