Semigroup and Prefix Computations on Improved Generalized Mesh-Connected Computers with Multiple Buses

Various augmenting mechanisms have been proposed to enhance the communication e ciency of mesh-connected computers (MCC's). One major approach is to add noncon gurable buses for improved broadcasting. A typical example is the mesh-connected computer with multiple buses (MMB). In this paper, we propose a new class of generalized MMB's, the improved generalized MMB's (IMMB's). Each processor in an IMMB is connected to exactly two buses. We show the power of IMMB's by considering semigroup and pre x computations. Speci cally, we show that semigroup and pre x computations on N operands, and data broadcasting all take O(logN) time on IMMB's. This is the rst O(logN) time algorithm for these problems on arrays with xed broadcasting buses.