7 巻 (2010) 20 号 p. 1520-1526
This letter proposes a multithreaded pattern matching algorithm which can efficiently distribute the patterns to be searched on multiple threads to achieve rapid pattern matching operation. The proposed idea is designed to fully exploit thread-level parallelism to enhance searching speed. By distributing a large number of patterns over multiple threads, pattern matching procedure experiences less cache misses and shows better performance. In addition, we propose to sort the target patterns according to the alphabetic order to achieve efficient data decomposition. From detailed experiments and performance analysis, our algorithm shows remarkable performance gain compared to the original Wu-Manber algorithm.