Second is a sad fact, that (I think) we're stucked in 70'ties if it concerns cpu architecture.
https://www.youtube.com/watch?v=sK-49uz3lGg
Neither is it Harvard or Neuman or different modifications - the base concepts are still the same.
We still use alu, registers, centralised synchronic "busses" and clock based processing in sequence (with a bunch of additional optmisation techniques like caches or out of order processing)
Is it CISC or RISC - all this is just deviations from one common aproach.
My thought was - it is miserable. Is it all that can be achived ?
Or we're just in hands of few monopolistic companies ?
Some hope is new RISC-V open source processor movement.
It is very nice fresh design - but it is still just opensourcing classical aproach.
Can't it be that there are other aproaches possible ?
IE distributed one with easy scalable subcomponents calculating one simple thing ?
Or: clockless one - working with smooth speed not in a rythm of "Paradeschritt" ?
Or: neural network like ?
Or: hierarchical one organized like LISP - to take over subsequent layers of compilation ?
Or: non-binary one - based on chips with more states/voltage than just 2 ?
Or ... ? Who knows what's yet in there ?
Why we are sitting still in the cave ? Just because of tons of code and systems writen ?
Could we recompile them to new hardware ?
RISC-V opened possibility of producing custom cpus using classicach architecture, I think it is not enough. There is a need to give people easy start for designing and simulating completely new aproaches.
makaronLab is (or rather I woud say may be) a tool to achieve this.