[d23]

Cool post! Thanks for submitting. I would never in a million years have thought to try out a polynomial fit for something like this. Unsurprisingly, it was slower, but still cool to even see you thinking that way at all.

[klyrs]

While I'm big on lookup tables for jobs like this, I'd have reached for bitmasks long before that polynomial...

[kardos]

Hard to compete with a 4-byte lookup table, but for decoding the 256-byte table can become: return ((dibit | (dibit -1))>>1) & 3;