基于SIMD指令集的SM2数字签名算法快速实现  

Fast Implementation of SM2 Digital Signature Algorithm Using SIMD Instructions

作　　者：韦薇 罗敏[1] 白野 彭聪 何德彪 WEI Wei;LUO Min;BAI Ye;PENG Cong;HE De-Biao(Key Laboratory of Aerospace Information Security and Trusted Computing,Ministry of Education,School of Cyber Science and Engineering,Wuhan University,Wuhan 430072,China)

机构地区：[1]武汉大学国家网络安全学院空天信息安全与可信计算教育部重点实验室,武汉430072

出　　处：《密码学报》2023年第4期720-736,共17页Journal of Cryptologic Research

基　　金：山东省重点研发计划(2020CXGC010115);国家自然科学基金(62172307,62202339,U21A20466)。

摘　　要：SM2数字签名算法是国家密码管理局发布的首个数字签名标准,已广泛应用于网上银行、电子政务等领域.本文提出一种基于高级矢量扩展指令集(advanced vector extension 512,AVX512)的SM2数字签名算法实现方案,有效提升了SM2数字签名算法的性能.结合单指令多数据集(single instruction multiple data,SIMD)运算特性,设计了一种新的冗余基数表示形式与数据排列方式,利用3比特冗余空间减少进位传播的次数,构建高效的并行素域运算模块.进而提出一种可变基点标量乘法的并行优化算法,在算法分支加入虚拟操作,按需存储点加与倍点的计算结果,结合底层数据表示形式消除8路分支的差异性.利用AVX512指令与分步点加方法加速固定基点标量乘法.签名与验签算法的性能比最新的SIMD实现分别提升了196%和69%.SM2 digital signature algorithm was published by the State Cryptography Administration of China.It has been widely used in online banking,e-governance,and other fields.This paper proposes an implementation of SM2 digital signature algorithm based on advanced vector extension 512(AVX512).Firstly,combining the characteristics of single instruction multiple data(SIMD),a reduced-radix representation and data arrangement are designed,so that the 3-bit redundant space can be used to reduce the number of carry propagation.Based on this,an efficient parallel prime field arithmetic algorithm is constructed.Then,a parallel optimization algorithm for variable base point scalar multiplication is proposed.Some dummy operations are added to the branches of the algorithm and the computation results of point addition and point double on demand are stored to eliminate the difference of the 8-way instances.Furthermore,a fixed-base scalar multiplication is accelerated using AVX512 instructions and a stepwise point addition.Experiments show that the proposed scheme can effectively improve the performance of SM2 digital signature algorithm.The performance of signature generation and verification is ahead 196%and 69%of the latest SIMD implementation.

关 键 词：SM2数字签名算法 SIMD指令集 AVX512 软件优化 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]