🚀 go-pugleaf

I belive that it is common knowledge that the H1 links are a new technology,
and offer substantially higher bandwidth.

I agree entirely about the communication/computation balance. My impression
is that market forces push Inmos to producing a processor with compute
power matching or exceeding the (sequential) competition. But for a
*parallel* machine, we all know that most applications are communication-
bound; most effort in programming is directed towards tuning communications.
I would rather see more silicon resources devoted to communication.

The other end of compute/communication spectrum is an SIMD array. A number
of us are trying to integrate both styles to get the best of both worlds!

Adrian Lawrence.   Oxford University.   ADRIAN @ uk.ac.oxford.vax