Speeding up the “Puzzle” Benchmark a “bit”

Abstract

We show how Baskett's "Puzzle" benchmark can be speeded up at least an order of magnitude by utilizing bit-vectors. Unlike many optimization techniques, the use of bit-vectors enhances the readability and understandability of the code. Since bit-vectors already utilize word-wide parallelism, it is unlikely that parallel processors will be able to solve the problem much faster.