Feels like a poor fit, given the limited number of cores available.
Would be awesome to see rp2350 or some such, where there are very low power io cores available that can do work whether the main core is on or not. Embedded really is one of the best places for many-core, but it's so so rare there are good offload architectures and puny Programmable IO systems.
Should out to folks like Silego/Dialog/Renesas with their GreenPAK; ultra tiny but interesting mixed signal little bits of programmable logic with a healthy dollop of peripherals!
Calling the RP2350's PIO units "low power io cores" is quite an exaggeration. Although they are technically turing-complete with a lot of hacking, they are absolutely awful at any kind of compute. Heck, you probably don't even want to let it handle UART parity calculation!
If anything the ESP32's Ultra-Low-Power Coprocessor would be perfect for such applications - but realistically it isn't worth the effort. Compute power usage is going to be negligible compared to what is needed for wifi and rotors, and running multiple realtime tasks on a single core isn't exactly rocket science either.
I agree that we have a wealth of capabilities available. The modern Cortex M7 cores are fantastically powerful.
Still, cramming everything into a main core had tradeoffs. It's a big core to wake up & run, with significant power draw. If you start running a bunch of stuff, it can be hard to meet realtime constraints, to provide reliable processing where needed.
The blanket statement that more cores is more complicated feels too broad. There's times when io cores are much simpler cinceptually, such as when you have realtime needs such as drones might. Subdivision of responsibility & dedicated cores is a powerful way to de-complect concerns. And it can bring huge energy savings.
Multi-core is also the chip making strategy that actually makes sense. We can cram so many transistors onto a tiny tiny chip, but what do we use the transistors for? We can keep trying to build bigger faster single cores, but there's only so much instruction level parallelism and caching we can actually effectively utilize. Having tiny io cores dedicated to specific tasks can use an incredibly tiny amount of extra die space, while providing incredibly low power dedicated processing for specific subsystems. Seems like a win.
Let's resist making broad statements like "you don't need". What's "needed" might not be what's best. We should have open minds to consider advantages & disadvantages. I for one think tiny io cores can later in amazing possibilities, have seen the rp2040 as a huge leap in capabilities (throwback to the Parallax Propeller with wide SMT like behavior) that's enabled a ton of interesting flexibility & novel low power embedded creations to get off the ground.
Would be awesome to see rp2350 or some such, where there are very low power io cores available that can do work whether the main core is on or not. Embedded really is one of the best places for many-core, but it's so so rare there are good offload architectures and puny Programmable IO systems.
Should out to folks like Silego/Dialog/Renesas with their GreenPAK; ultra tiny but interesting mixed signal little bits of programmable logic with a healthy dollop of peripherals!