摘要
RSA算法的核心运算为模幂运算,具有极高的计算复杂度和计算密度。研究了基于GPU的RSA加解密方法,根据算法并行度的不同,采用了串行和并行两种方法进行实现。实验结果表明,基于GPU的RSA加解密方法成功地实现了硬件加速。相对于CPU上运行的RSA加解密方法,该算法极大地提高了执行效率,获得了最高4倍的加速;在处理数据较少时,并行算法比串行算法有优势,但是串行算法在处理数据较多时性能逐渐增加,并获得了最高性能。
GPU that has powerful massive data parallel computing is applied widely in the computing intensive area. The core of RSA algorithm is a modulur algorithm which has high computing complexity. In this paper, the encryption and decryption of RSA based on GPU was studied. According to the different parallel level, serial algorithm and parallel algorithm were implemented. The result shows that the method of RSA based on GPU, which is more efficient compared with the RSA algorithm implemented on CPU, can get the acceleration of more than four times. When dealing with little data, parallel algorithm has a better performance than serial algorithm. However with the input data increasing, the performance of serial algorithm exceed the parallel one gradually, and get the best performance at last.
出处
《计算机应用》
CSCD
北大核心
2013年第A01期35-39,共5页
journal of Computer Applications
基金
全军军事学研究生课题(2011JY002-435)
湖北省自然科学基金资助项目(2010CDB01501)
关键词
统一计算设备架构
图形处理单元
模幂运算
并行
Compute Unified Device Architecture（CUDA）
Graphic Processing Unit（GPU）
modular exponentiation
parallelization
