I-Corps: Harnessing Unary Computing for Modern Applications

University Of Minnesota-Twin Cities, Minneapolis MN

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is the development of a novel computing method called “unary computing,” which makes computations very efficient with limited hardware resources to perform complex calculations. Applications of this method include signal processing (image/video/audio/radar) and machine learning for autonomous vehicles, UAVs, satellites, as well as cloud computing and internet-of-things (IoT) systems. The proposed methodology has the potential to significantly reduce power and energy requirements for emerging applications on the cloud and at the embedded-systems level, benefiting chip manufacturers and cloud platform companies. Furthermore, IoT applications that may benefit from reduced hardware cost and smaller batteries include medical implants, ultra-low latency trading, and unmanned aerial vehicles (UAVs)/ This I-Corps project is based on the development of unary computing, which uses an “uncompressed” data representation (unary) instead of the traditional binary representation to make computations very efficient. Conventional methods would require many multiplications and addition operations to calculate functions such as cosh(x), tanh(x) and y=x^0.45 through polynomial expansions. Unary computing is able to perform these calculations cheaply and directly without any multiplication operations or additions. The proposed method shows remarkable performance benefits compared to conventional binary computations. For example, when implementing tanh(x) at 8 bits of resolution, a hybrid unary-binary method can implement the function with only 24 units of area on FPGAs (a particular type of integrated circuits), whereas conventional binary would need 509 units. The unary method outperforms binary in other performance metrics, such as speed and power. The method results in approximately 20%-30% area improvement for constant-coefficient multiplication, which is heavily used in machine learning applications. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.