Then again, the cores on RP2350 run so fast and efficient, there's no point trying to "offload" the CPU. I can probably get the repacking down to 1-2 cycles per bit. Even the shittiest C loop is taking less than 1% of max performance.

Just hoped maybe I could find a _neat trick_ to get this for free. Wasting that memory bandwidth will probably end up more power than just using the CPU.