knudsen@ihwpt.UUCP (mike knudsen) (03/01/86)
I would like to hear from anyone who is either working with (or has worked with) a system using the INMOS Transputer chipset, or has at least considered the system for their project and has rejected it for good reasons or has not yet finalized the decision and gotten underway. There's been some discussion of transputers on these groups but no real experience that I could see. I've heard stuff like "like the Intel 432 -- fascinating stuff but nobody's using them." I'd like to hear from some "nobodies" who can disprove that. I think this technology bears looking into. Just to stir up the pot: much of the transputer concept is (coincidentally) exactly like the network of digital signal processor (DSP) microchips I built a few years ago and reported on in the 1983 ICASSP (Intl Conference on Acoustics, Speech, & Signal Processing) Proceedings (see Knudsen, M.J., "MUSEC: ..."). Both schemes share the notions of several processors with local memories (no global memory) connected via hi-speed serial links, each doing a *different* subtask of the application (MI-MD architecture, in parallel- processing terms). INMOS and I claim many of the same advantages for our systems. INMOS' major contribution in my opinion is the unified means of programming the whole mess in transparent fashion using the Occam language. I'd like to hear more about that too, but that's another newsgroup. Thanks, mike k