TX-RSA: A High Performance RSA Implementation Scheme on NVIDIA Tegra X2

Driven by computer vision and autopilot industries, embedded graphics processing units (GPUs) are now rapidly achieving extraordinary computing power, such NVIDIA Tegra K1/X1/X2, which are widely used in embedded environments such as mobile phones, game console and vehicle-mounted systems. Such performance advantages give embedded GPUs the possibility of accelerating cryptography that also requires high-density computing. In this paper, we implement TX-RSA in embedded GPU platforms, i.e., NVIDIA TX2, to accelerate the most prevailing public-key cryptosystem, RSA. Various optimization methods are employed to promote the efficiency, including multi-threaded Montgomery multiplication and CRT implementation on the resource-constricted embedded GPUs. Within 20 W of power consumption, TX-RSA can deliver 6,423 ops/s of RSA encryption and 131,324 ops/s of RSA decryption, which outperforms implementations in the desktop GPUs and embedded CPUs in the perspective of performance-to-power ratio.