''Scalable Basic Algorithms on a Linear Array with a Reconfigurable Pipelined Bus System''

Jerry L. Trahan

Yi Pan

Ramachandran Vaidyanathan

Anu G. Bourgeois

Proc. 10th ISCA Int'l. Conf. Parallel and Distributed Computing Systems, (Int'l. Society for Computers and Their Applications, New Orleans, LA, Oct. 1997), pp. 564-569


The Linear Array with a Reconfigurable Pipelined Bus System (LARPBS) is a parallel computational model based on current optical technology. Many practical and efficient algorithms have been designed for the LARPBS. Most algorithms, however, assume that the problem size determines the physical size of the model. As is the case for other reconfigurable models, algorithmic scalability of the LARPBS poses several difficulties. This paper studies the scalability of basic data movement operations on the LARPBS to efficiently implement arbitrary sized algorithms on a given sized LARPBS. We show that many algorithms are either completely or ``relatively'' scalable. The results extend the use of many existing LARPBS algorithms employing these operations.