The local similarity problem is to determine the similar regions within two given sequences. We recently developed a dynamic programming algorithm for the local similarity problem that requires only space proportional to the sum of the two sequence lengths, whereas earlier methods use space proportional to the product of the lengths. In this paper, we describe how to parallelize the new algorithm and present results of experimental studies on an Intel hypercube. The parallel method provides rapid, high-resolution alignments for users of our software toolkit for pairwise sequence comparison, as illustrated here by a comparison of the chloroplast genomes of tobacco and liverwort.