基于精英策略的遗传算法在功能验证中的应用

Genetic Algorithm Used in Functional Verification Based on Elite Strategy

针对集成电路功能验证中覆盖率收敛较慢的问题,通过分析简单遗传算法(SGA)中精英个体的特征,提出了一种应用于功能验证的精英策略。将本代优秀个体和本代适应度高的历史优秀个体视为精英个体,给予额外交叉机会。基于本文策略的精英遗传算法(EGA)可得到覆盖率广、重复性低的验证向量,缩短功能验证的时间。采用互相关函数的硬件计算单元作为验证模型,在Matlab中模拟功能验证的过程,实验结果表明:与SGA相比,EGA使验证时间缩短了14.8%,功能覆盖率从93%提高到95%,有效地提高了功能验证效率。

This paper addresses the problem of slow convergence occurring in integrated circuit functional verification. By analyzing the features of elites in simple genetic algorithm （SGA）, this paper proposes an elite strategy for functional coverage, in which elites are selected in the qualified group and crossed by normal group. Elite genetic algorithm （EGA） based on elite strategy may generate verification vectors with high coverage and low reproducibility such that the functional verification can be accelerated. By adopting an NCC-computing cell as the verification model, the simulation on the function verification process in Matlab is made, whose results show that compared with SGA, EGA can reduce simulation time for 14. 8 % and functional coverage increase from 93% to 95%. The efficiency of integrated circuit functional verification is effectively raised.