Pipelined architecture for computational nanotechnology

Abstract

Nanomechanical computing elements which are scalable in terms of input size and depth of propagation path are analyzed using a bounded continuum model. Boolean logic functions of NOT, AND, OR, and XOR are realized using helical latch, reset spring, and translating rod assemblies. Building upon these components a design for two-level logic operations is presented. The helical latching mechanism calculates the Boolean output function as a positional displacement from a known reset state, which occurs exactly once during each instruction cycle. To balance forces a symmetrical rotor is used to counteract applied forces by replicating input rods. This has the beneficial side-effect of providing intrinsic fault-detection capability within a gate and also decreases the rotation required for a full cycle from 360 degrees to 180 degrees. This design is further enhanced to allow operations of arbitrary word length by subdividing the logic disc into sectors where each sector contains all the components necessary to operate on a single bit. The benefits of increasing the disc diameter needed for additional bits include a further reduction in disc cycle rotation as a result of subdividing the disc into sectors. Since the inputs are sampled sequentially, throughput of resultants can be increased directly by pipelining multiple bit operands. For n inputs per logic gate, the maximum speedup for a single level of logic is (n+2). Generally, speedup is bounded by (n+2)/p where p denotes the number of cycles between initiations of the pipe.