Newsgroups: comp.sys.transputer From: Charles Sweeney Reply-To: CharlesSweeney@compuserve.com Subject: Re: A Problem ? Organization: Embedded Solutions Ltd Date: Fri, 07 Nov 1997 11:45:44 +0000 Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <3462FF20.7AE1@compuserve.com> Jonathan Bromley wrote: > Now, maybe our hype is carrying us away just a teeny little bit, > isn't it? Handel-C is a seriously good idea with a fantastic future, > and we know you have to harp on about its C-like syntax so that all > those sad people out there who believe (programming==C) will at least > look at it. But you and I know Handel-C is really occam in disguise, > and this newsgroup of all places is surely somewhere you can not > only acknowledge that but also be proud of it! A fundamental feature of Handel-C is its parallelism which is based on the same CSP ideas as occam: the 'par' construct and channel input and output. However there are also many differences from occam associated with hardware as a target eg bit width of variables specified by programmer, each assignment takes exactly one clock cycle, variables can be assigned to FPGA pins. Also the syntax of C is used wherever possible because there are more people in the world who are familiar with C than with occam or most other languages in the embedded systems market. We don't try to hide any resemblance to occam, it's just that we don't feel any need to flaunt it, we do mention CSP in our documentation. From many years of programming in occam off and on from 1987 I have found it to be excellent, particularly the checking which helps ensure correct operation of the program. Handel-C is founded on the same mathematical principles (CSP) to ensure correct operation and reliability which is becoming more and more important as programs get larger and larger. > > BTW, occam/Transputer was a seriously good idea with a fantastic > future, too. But it folded commercially. PLEASE, for the sake > of us who like to be able to _reason_ about our hardware and software > as well as just getting it to work, make Handel-C useful enough so > that it won't fold commercially. That means, amongst other things, > making it possible to instantiate Handel-C modules within Verilog, > VHDL and schematic designs; and vice-versa. Yes, we're very aware of this need. But many applications can be programmed completely in Handel-C. > > Frustrated occam programmers should note the above with some care. > Many of the things that we thought Transputers would be good at > 10+ years ago now turn out to be ideal applications for reprogrammable > hardware programmed using a CSP-based hardware description language. > With 10^5 gates per FPGA available today and 10^6 gates just round > the corner, you can do serious work. It may well be that you will > be employed as a hardware writer, very soon. Well said!!! We feel that occam programmers (me included!) can adapt to Handel-C very easily because of their mind set of thinking in terms of parallel processes. FPGAs are a massively growing market and are already being used for DSP type of applications (programmed in VHDL etc). So we are using a chip which will not be scrapped, will increase rapidly in power, has multiple competing sources, is made in the US so is acceptable there - compare all this with the transputer! We have a much simpler task than Inmos ever did because all we're doing is providing a much simpler way of programming FPGAs than existing tools. Regards, Charles Charles Sweeney, Engineering Director, Embedded Solutions Ltd Tel/fax +44 1235 510456 Email CharlesSweeney@compuserve.com or csweeney@embedded-solutions.ltd.uk 6 Main Road, East Hagbourne, Didcot, Oxfordshire. OX11 9LJ. UK.