Efficient CRCW PRAM Emulation on Practical Networks

Abstract

A new interconnection network is proposed for the construction of massively parallel computers. The systematic construction of this network, denoted RCNFULL, is performed by methodically connecting together a number of basic atoms where a basic atom is a set of fully connected nodes. Key communication characteristics and efficient routing algorithms are derivedfor RCN-FULL. An 0(log(N)) sorting algorithm is shown for RCN-FULL and RCN-FULL is proven to deterministically emulate the CRCW PRAM model, with only O(log{N)) degradation in time performance. Finally, the hardware cost for the RCNFULL is estimated as a function of its pin limitations and compared favorably to that of the hypercube.