Efficient Function Mapping in Nanoscale Crossbar Architecture

Nanoscale crossbar architectures have been proposed as viable alternatives for overcoming the fundamental physical limitations of CMOS technology. However due to the manufacturing processes for Nan fabrication and their smaller feature sizes, defect densities are higher. This paper presents an efficient function mapping method in the presence of high defect rates for nanoscale crossbar arrays. Given a function and a defect map that describes fault patterns in the crossbar architecture, the approach described here tries to find a valid function mapping, if one exists, using a matrix representation. A set of constraints are derived to preserve semantics and then Integer Linear Programming (ILP) is used to solve the equations. Experimental results show the proposed approach provides efficient utilization of nanoscale crossbars in mapping functions in presence of high defect rates.