A High Throughput SM2 Digital Signature Computing Scheme Based on Graphics Processing Unit Platform

With the pervasiveness of secure data transmission techniques and increasing requirements of information authentication, the public key-based digital signature scheme has been extensively used in various fields. However, the process speed of digital signature has gradually become the bottleneck of various security and high-concurrency applications. In this paper, a high-throughput SM2 digital signature computing scheme based on Graphics Processing Unit(GPU) platform is proposed. Firstly, the basic operations are optimized by low- level instructions of GPU. Then, according to the characteristics of GPU platform, the addition chain of SM2 recommended prime number is reduced and the speed of modular inverse operation based on Fermat's theorem is improved. Furthermore, a pre-computing table is constructed and the repeated doubling algorithm is introduced to accelerate the unknown point multiplication. Due to the construction of pre-computing table, divergence of threads can be successfully avoided. The experiments show that the proposed scheme can effectively speed up SM2 algorithm, and the throughput of signing and verification can respectively reach 76.09 million ops and 3.46 million ops on RTX3090.