Techniques for data-parallel searching for duplicate elements
Brenton J. Lessley, Kenneth Moreland, Matthew Larsen, Hank Childs · 2017
We study effective shared-memory, data-parallel techniques for searching for duplicate elements. We consider several data-parallel approaches, and how hash function, machine architecture, and data set can affect performance. We conclude that most choices of algorithm and hash function are problematic for general usage. However, we demonstrate that the choice of the Hash-Fight algorithm with the FNV1a hash function has consistently good performance over all configurations.