Newsgroups: comp.sys.transputer
From: davidb@ndl.co.uk (David Boreham)
Subject: Re: Transputer VME board?
Organization: Network Designers Limited
Date: Mon, 14 Mar 1994 10:14:15 GMT
Message-ID: <CMnGFr.1qC@ndl.co.uk>

In article <1994Mar10.163930.21772@eso.org>, rreiss@eso.org (Roland Reiss) says:
>
>In article 17072@sun0.urz.uni-heidelberg.de, Martin Feuerstack <feuerstack@physi.uni-heidelberg.de> writes:
>>I'm in search for a transputer VMEbus master. I already have Parsytec's 
[snip]
>>Does anybody know such a board?
>>
>Try Inmos' IMS B016-4 which has 16 MByte dual-ported RAM and a T801 with 256kByte
>private RAM. Don't know how the Parsytec reset scheme looks like, the B016 offers
>4 link interfaces with 4 independent service signal ports, so you can connect up to 4 transputer systems (similar to the BBK-S4 I think).
>

I believe that the "Parsytec" reset scheme involves a "services" connection for every
link and is bidirectional. That is, any adjacent node may reset it's neighbour. Every
node has visibility of the error status of its neighbours. This is a "reset graph".
Sorry If I've got this wrong.

The B016 implements the "B004/B003" reset scheme, where there is a reset hierarchy.
The reset topology is a tree, with the intention that many processor nodes may be
grouped together at the same level in the tree. In this case, the resets are in parallel and the
errors are OR'ed together. In the B004, there is only one "subsystem" port and therefore
the reset tree is a binary tree. After much discussion, I put four subsystems on the B016
because then it can emulate four B004's, giving independent control of four sub-networks
of processors.

So, there are fundamental differences between the two schemes. Whether you are able to
use the B016 will depend upon your application's adaptability to the INMOS reset scheme.

