Newsgroups: comp.sys.transputer
From: jdd@aiki.demon.co.uk (Jim Dixon)
Subject: Re: C004 links witch speed?
Organization: AIKI Parallel Systems Ltd
Date: Fri, 29 Jul 1994 16:30:52 +0000
Message-ID: <775499452snz@aiki.demon.co.uk>

In article <31alje$gfl@liberator.et.tudelft.nl>
           ronald@dutepp13.et.tudelft.nl "Ronald in 't veld" writes:

> I'm working on a project with lots of transputers connected by means of a
> C004 link switch.  Since I will use the switch dynamically (the switch will
> have to be reprogrammed during program execution) I need to know:
> 
> How fast does this thing work??????????
> - What is the setup time and execution time for one link reprogram?
> - After what time is the link ready for use after reprogramming
> - What is the total time for 16 links to be reprogrammed.
> 
> I have used the estimate of 200 us. for a total reprogram so far... How 
> realistic is this estimate?    Class, anyone?

As I suppose you must know, the C004 is programmed by sending conventional
messages to it.  If you are going to reprogram 16 links, you need to
send 16 messages like
    BYTE 1 ; BYTE link1 ; BYTE link 2
That is, a total of 48 bytes.  At roughly 2 bytes per us, this should
take 24 us.  Allow 1 us to set it up, and you have 25 us.

This assumes that you transmit the configuration messages at one go.
If you send 16 separate messages, add about 16 us, so around 41 us.

The three-byte configuration messages arrive over a period of 1.5 us.
Dunno what you are used to, but at the chip level that's 1500 nanoseconds,
a long long time.  The lag between a link being reprogrammed at its being
ready for use is not documented in the databooks, but it certainly
should not exceed a microsecond.
-- 
+-----------------------------------+--------------------------------------+
|  Jim Dixon<jdd@aiki.demon.co.uk>  |	    Compuserve: 100114,1027	   |
|AIKI Parallel Systems Ltd + parallel processing hardware & software design|
|	     voice +44 272 291 316  | fax +44 272 272 015		   |
+-----------------------------------+--------------------------------------+

