Fast vector arithmetic over $\mathbb{F}_3$

Abstract

We show how binary machine instructions can be used to implement fast vector operations over the finite field $\mathbb{F}_3$. Apart from the standard operations of addition, subtraction and dot product, we also consider combined addition and subtraction, weight, Hamming distance, and iteration over all vectors of a given length. Tests show that our implementation can be as much as 10 times faster than the standard method of using modular arithmetic on arrays of bytes. For computing the Hamming distance even a factor of 33 can sometimes be reached, provided a recent CPU is used.