CCT: a new VLSI architecture for parallel processing

Abstract

We propose a VLSI implementable architecture called Cube Connected Tree having advantages of both trees and hypercubes. This structure has fixed low degree nodes for any size of network, unlike hypercubes, where the node degree is dependent on the size of the hypercube. Complexity of VLSI layout of this structure has been addressed within the grid model of C.D. Thompson (1984). By using spare links and PE's, fault-tolerance capabilities of the system has been enhanced. Easy programmability of this structure has been demonstrated by designing polyalgorithmic algorithms for sorting and discrete Fourier transform.