🚀 go-pugleaf

Subject: Semantics of In and Out instructions for memory channels?

Dan McCabe <mccabe%unc.uucp%thorin@mcnc.org> writes to this mailing list:

> Does anyone out in Netland know details about the semantics of the In and Out
> instructions when a memory channel is specified?  
> I know that a channel is initialized with NotProcess.p (-1).  I presume
> that a process descriptor is kept in the channel variable.  Is the
> channel a linked list of waiting processes?  
                                                No. It is either the process
descriptor or it is NotProcess.

>                                              It seems to me that this would
> have to be the case, since several processes may Out to the same channel before
> another process does an In.  
                                No. Only 1 process may Out. Channels are 
1-1 and implement synchronised communication. The (only) process which 
is outputting will be dscheduled when it performs its output if no process
has yet performed an input. No further output can occur because the only
process which is using the channel for output is descheduled. When the 
matching input occurs the outputting process is rescheduled.

> Are the registers saved in the workspace if the instruction blocks
> (when Out executes and there is no pending In or vice versa)?  

The only values in registers when an input or output instrcution is executed
are the channel id, data pointer and message count. The pointer is saved into
the workspace as is the instruction pointer.  Nothing else is saved.

>                                                             If you
> treat the list of processes as a queue, you also need a tail ptr for
> efficient insertion (as is the case for general process scheduling). 

True. There is a back pointer for the scheduling queue; there are no
process queues for channels.

> Is there some encoding that says that the channel is waiting either for
> an In instruction or for and Out instruction? 
                                                No. It is up to the user
to ensure the channels are used correctly. When a suitable programming 
language (eg occam) is used this can be (and is) be done by a compiler.

There is a good discussion of this in ``Communicating process architecture'',
published by Prentice-Hall international, ISBN 0-13-629320. Chapter 2,
The transputer implementation of occam describes the implementation and
gives a rationale for the communication model.

``... commonication is point-to-point, synchronused and unbuffered. As a
  result, a channel needs no process queue, no message queue and no message
  buffer.''


Roger Shepherd, INMOS Ltd   JANET:    roger@uk.co.inmos 
1000 Aztec West             UUCP:     ukc!inmos!roger or uunet!inmos-c!roger
Almondsbury                 INTERNET: roger@inmos.com
+44 454 616616              ROW:      roger@inmos.com OR roger@inmos.co.uk