Not sure if twitter threads are well-received here, let me know if this is unwelcome.
Thanks, interesting. Here are unrolled threads for convenience:
Interesting note in there:
I’ve got a tabletop Tektronix SS7 analyzer here that features a whopping 7 transputer cores on two enormous cards, used for analyzing telephone system traffic
Also this is interesting:
ISDN hardware saw a lot of transputers being used for example
I know Motion Media’s ISDN-connected videophone was based on transputer technology - but I think it might have used the last-gasp ST20 which was a microcontroller repackaging of transputer technology. It would still be programmed in the CSP paradigm - Communicating Sequential Processes - but those processes would all be running on the same single core, using the inbuilt hardware scheduler.
Among the places that Inmos got into trouble, in my view:
- getting only half the requested bootstrap funding because of a change of government
- pushing Occam for too long before supporting C
- expecting CSP to be an acceptable paradigm… as cores got faster and shared memory became the norm, passing messages over external links became a bottleneck
- failing to scale up performance, in part because a stack machine is difficult to speed up
- losing all their chip design experience by taking too long and not retaining people, replacing with inexperienced engineers and managers
- being in the UK where engineers are cheap but venture capital all but non-existent