For one thing, the PRUs require TI’s bloated mess called CCS and knowing C whereas the PIO assembly can be written in Python. What’s really innovative is the ability to get PRU-like functionality on a literal $1 IC. Sure, you only get 32 instructions per state machine, but not everyone needs full blown ARM cores for real-time.
They don't require it. When I used it, everything was just written in assembly (you get a few more instructions but you're still basically doing bitbanging with cycle counting, so using C is a bit awkward anyway). The PIOs are a pretty neat trimmed down version of the idea though, clearly someone put a fair amount of effort into making them as flexible as they are for how minimalist they are.
I think C is a trap, honestly. I have used the PRUs before and wrote my code in C. 99% of the time, I was just reading the assembly outputs trying to figure out how many instructions some C statement actually compiles to. Should have just written raw assembly for the timing-sensitive parts.
