FUNSC: A GUI Software for Stochastic Computing

Stochastic computing (SC) is an unconventional approach for performing computation by logic circuits, where data is represented and processed in the form of pseudorandom bit-streams. Due to its main advantages, i.e., low-area hardware and fault tolerance, SC is emerging in several applications including image processing and neural networks. The design process for SC, however, is challenging. This paper introduces FUNSC, a design automation software tool, which efficiently generates SC circuits for computing mathematical functions. The tool features a user-friendly graphical user interface (GUI) that accepts input functions in their mathematical expressions and allows users to adjust the hardware complexity versus computational accuracy trade-off. To generate SC circuits for a desired function, FUNSC employs a two-step process: first, it approximates the function using a polynomial derived from Maclaurin series expansion, and then it applies Horner’s Rule or Double-NAND Expansion rearrangements to map the polynomial to an SC circuit and displays generated schematic in the GUI. Our results demonstrate that the generated SC circuits achieve low error rates while successfully computing eligible functions.