Emulation of hypercube architecture on nearest-neighbor mesh-connected processing elements

The problem of emulating a hypercube architecture on nearest-neighbor mesh (NNM) connected processing elements is addressed. The motivation for this study was the recent promotion of the transputer as a possibly very powerful building block of large highly parallel, scalable computer architectures. It is an attempt to achieve hypercubes of different dimensions with a constant, small number of links per processing element. The idea was investigated by means of some less abstract ways of reasoning and computer simulation in an attempt to find the best form of NNM connected processing element structures form the viewpoint of the optimization of certain parameters of the emulated hypercube.