GPU Acceleration of Longest Common Substrings Algorithm

Abstract

The Longest Common Substring of two strings is a character sequence that appears in both texts and is the longest of these. The method is widely used in several text similarity measurement methods, usually used multiple times on the same textual data. There are several already known methods to solve the problem, but these are mostly based on very time and memory intensive procedures. This paper presents a novel data-parallel model to solve the same problem, available for GPU implementation. As our experimental results show, the data-parallel implementation is significantly faster for long textual data.