Performance Evaluation and Design Optimization for Flexible Multiple Instruction Multiple Data Elliptic Curve Cryptography Crypto Architecture

摘要

The parallelism and granularity of a Multiple Instruction Multiple Data (MIMD) crypto architecture is evaluated in this paper, with a flexible and efficient implementation of scalar point multiplication over GF(2~m) operands. Unlike the related implementations in the literature that do not combine all such features, the studied MIMD architecture in this paper is modular, scalable, supports arbitrary operand sizes, and is capable of handling different parallel operations. An exhaustive simulation-based study of the different system-level performance aspects of this architecture is studied under several input data patterns. Simulation results reveal how such a flexible architecture significantly increases the level of parallelism, and hence, the utilization and throughput at a slight cost of increased area for the task controllers. Results also confirm the superior timing characteristics of the MIMD as compared to other existing architectures. Furthermore, the impact of different scheduling schemes on the performance of the MIMD architecture is investigated. A simulation tool is developed to allow the designer not only to visualize and evaluate the behavior of the MIMD architecture, but also to obtain the parameters of the architecture that achieves the optimal degree of parallelism as well as granularity without compromising area, power, and timing performances.