[londons_explore]

Very efficient for storage and memory bandwidth, but such a scheme is a headache for high throughput hardware implementations (at least compared to regular 4 bit math, which can be packed really really densely)