Sieving using bucket sort

Abstract

This paper proposes a new sieving algorithm that employs a bucket sort as a part of a factoring algorithm such as the number field sieve. The sieving step requires an enormous number of memory updates; however, these updates usually cause cache hit misses. The proposed algorithm dramatically reduces the number of cache hit misses when the size of the sieving region is roughly less than the square of the cache size, and the memory updates are several times faster than the straightforward implementation. © International Association for Cryptologic Research 2004.