Thanks for the kind words and great feedback!
Yeah I remember AltiVec was all the rage back in the G4 days, but I never learned how it worked. I don't know what specifically made the instructions nicer than SSE. How does AltiVec compare to current x86 SIMD instructions you think?
The PS3 cell experiment is definitely something I wish I knew more about.
At first glance it sounds like such a great idea. Make the stuff that is less important run slower and the really demanding stuff run fast with specialized processors.
But I guess as you say, they got the balance wrong. There where perhaps more general purpose code you needed to run fast.
But I also want to speculate that this was simply a too early time for hetereogneous computing. And maybe Sony had their software strategy wrong.
I mean if you look at Apple e.g. they stuff all their specialized hardware behind frameworks that you use. Developers don't relaly have to know much about the details. But Sony seem to have just dumped all the complexity on the developers: "Here figure this out yourself!"
It would have been interesting to see what a company that gets both hardware and software could achieve with a PS3 style solution for a game console or desktop computer.
I mean the M1 seems to be a lot of the ideas of PS3 with all its specialized hardware. OTOH, I guess the devil is in the detail. M1 also has strong general purpose CPUs.