Extracting a Longest Common Subsequence by a Systolic Algorithm on Multicomputers with Fewer Processors

Abstract

A longest common subsequence (LCS) of two strings with mn/log n processing elements (PEs) an efficient is a common subsequence of the two strings of maximal length. The LCS problem is to find an LCS of two given algorithm runs in O(log m + log2 n) time [17]. On the strings and the length of the LCS. The LCS problem has same model, a cost-optimal algorithm uses been the subject of much research because it can be mn/max{log m , log2n loglog n} PEs and takes applied to many areas. In this paper, a scalable systolic O(log m + log2 n loglog n) time [17]. Because it is algorithm is presented. For two given strings of length infeasible to build a PRAM with more than a few tens of m and n, where m ≥ n, the systolic algorithm can PEs, the PRAM models are not practical for this problem solve the LCS problem in m + 2r – 1 (respectively n + when m and n are large. 2r – 1) time steps with r < n/2 (respectively r < m/2) In contrast, the systolic model [10] is practical for the processors. Experimental results show that the algorithm LCS problem. A systolic architecture is a regular array of has satisfactory performance on the IBM SP2 identical PEs. It can be implemented with VLSI chips or multicomputer.