The Interconnection Problem

The engineering research of the next decade will most likely bring to fruition the integration of different functionalities on a single small platform, such as compact electro-photonic modules using engineered photonic bandgap structures to reduce the size of optical modulation-demodulation elements. I expect that the newest integration architectures will necessarily include fully three-dimensional circuits and up to a million or so "zero cost" single devices. These integrated modules will, in themselves, be complex networks of individual elements, each element type described by a statistical distribution of characteristics. We will need a breakthrough in the methods of integrated circuit simulation and design in order to deal with the complexity of designing these modules and to deal with the latency of signals travelling across long paths or through many connections. Right now, the simulation and design software for merely pure electronic integrated circuits, assuming that each element type is identical, is a major bottleneck in the production of application-specific integrated circuits (ASICS). One possibility in the far future is to harness the methods of directed self-assembly of the network of devices, much as our brain manages to learn from its environment how to assemble the synapses between neurons. We are not even close today!

0 0

Post a comment