Newsgroups: comp.sys.transputer From: Paul Walker Subject: Re: Transputer - alternate source? Organization: 4Links for technical help Date: Mon, 2 Feb 1998 19:20:43 +0000 Mime-Version: 1.0 Message-ID: In article <886437490.465174534@dejanews.com>, Brian.Oneill@ntu.ac.uk writes >I see the transputer equivalent as an attractive solution for embedded >systems applications. Other more high level applications will find >standard operating systems more attractive (e.g. RISC-OS for the ARM), >which offer greater functionality at the expense of performance. > >One problem will be defining when a system is a equivalent or not. My >own view is that to claim equivalent status the processor must have a >light weight kernel and easy to use, low latency, inter-processor >communications. Agreed to both of these. The Kroc kernel is a very light-weight implementation of the transputer's microcoded comms and sheduling instructions. So also is Eonic's Virtuoso RTOS, which allows the user to choose the language while keeping the CSP model. Both run on a variety of processors. On the issue of an MMU, I'd prefer not to have one, but people are familiar with them and leaving it off could make an unnecessary disincentive for people to buy. Perhaps the best approach is to follow the favourite manufacturers. ARM's recent ARM 9 cuts a load of complexity and functionality from its MMU, so once again we can see things heading in the direction of the transputer. Elsewhere, Brian suggests we won't see another single chip device. We certainly need to start with at least a two-chip set of processor + (probably FPGA) comms. Brian is probably further down this road than anyone for OS. DSS are further down the road for DS, except they need the dual-port RAM which takes up more space than some users can afford. But there really are advantages in a single chip, and the chip-set should be seen as a production prototype for the single chip. It's heretical to suggest anything other than OS and DS. There are alternatives, like 1355 HS, and like putting the packet protocols on top of a familiar PHY such as Ethernet. Horrors in some respects, but with the IP people talking about tag switching and label switching, they are moving in our direction. It might be rather nice to offer them the next generation of Ethernet circuits! It might also be wise to move away from any IP (different sort) owned by ST. The customers who have been let down need something that works with their existing kit, so the heresy is probably better not followed. I'd be interested in their views. Paul -- Paul Walker 4Links phone/fax paul@walker.demon.co.uk P O Box 816, Two Mile Ash +44 1908 http://www.walker.demon.co.uk Milton Keynes MK8 8NS, UK 566253