Thu Aug 16 17:11:54 EDT 2018

Why do FPGAs use LUTs?

I don't know, but a guess is that a LUT doesn't introduce any bias
against certain functions.  Each n-ary function is implemented with
the same cost.  As such, LUTs are universal.

They are probably also easy to implement as compared to other

The question is more: why 3 to 1, and not any other n to m

There does seem to be some variability here.  3-1, 4-1, even mult-out?

This says 4-1 is best for size, 5-1 is best for performance.