NMOS Transputer

SK8 4 U

Hi all...

I am back from a week at the lake all rested and ready to ask yet another Q about an obscure computer...

Does any one remember the NMOS transputer chips or the boards that were available to install in a PC?· They were serially interconnected u-processors that could be connected in highly parallel archetecture.· NMOS has long since bit the dust, but I think that some other company bought their patents and foundries.

All the best,

73

Joe

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cats are the creator's method of demonstrating to mankind that man is not the supreme being.

SteveW

Yep - it's inmos, not nmos (or rather, it was inmos).
SGS-Thomson (ST) bought the rights to the transputer architecture - it's still doing good service in set-top boxes, although it's generally programmed in C, rather than occam, which takes all the fun out of it.
The concept of a fast, serially linked processor lives on, though, in DSPs from TI and AD, and there's still a fair amount of interest in occam (including some very usable PC implementations). This will only increase, as multi-core CPUs get more common, and interprocess communication becomes more and more important, along with partitioning software among CPUs.
(occam for linux is free from my old university - yay!)
http://www.wotug.org/kroc/
Lots of tutorials and teaching stuff too, since UKC uses occam as a parallel processing teaching language.

Steve

SK8 4 U

Hi Steve...

I can remember remember reading about a hypercube setup using the transputer chips and a (then) state of the art IBM AT to house the cards in. The AT was simply the terminal for the Transputer boards, and provided power as well to them. I vaugely remember someone at MIT who set up a 16 node hypercube to calculate planetary orbits...it ran much faster than the somewhat available PDP 11/???, and obviously much faster than any PC could at the time. Now, I suppose my old outdate PII that I play with linux on could probably run faster with a poorly coded program, but hey, at the time it was an amazing machine. This was at a time when there were "mathematic accelerator" boards for the AT that had zillions of chips on them, and of course they were thousands of dollars. This was when computing was still interesting and in my oppinion fun...something that the latest Athelon or whatever, as impressive as they may be, just can't duplicate.

Maybe I am jaded...

All the best,
73
Joe

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cats are the creator's method of demonstrating to mankind that man is not the supreme being.

pharseid

· I actually programmed on some of those transputer boards years ago. In C. There was a library for using the communication channels. I never used Occam because I felt I would be porting my code to other platforms, so I really don't know much about it. I read Communicating Sequential Processes, and Occam seemed to follow that sort of philosophy.

· As has been mentioned, a couple of DSP's have been built for parallel processing and I've read in a couple of places that some guy has an HDL model for the transputer (I can't remember if it was Verilog or VHDL).

· I think there are still interesting things going on, but the world is so PC-centric that you don't hear about them.

-phar