基于遗传算法的功能覆盖率收敛技术

Functional coverage convergence technique based on genetic algorithm

针对集成电路验证向量生成与功能覆盖率收敛的问题,提出一种基于遗传算法的功能覆盖率收敛技术.通过计算分析遗传算法中遗传算子的概率分布函数,获得由比例选择算子、均匀交叉算子以及二元变异算子组成的遗传算法,得到覆盖率广、重复性低的验证向量,在最短仿真时间内达到预先设定的功能覆盖率.实验采用基于Turbo芯片的图像处理硬件加速器作为验证模型,将遗传算法嵌入到以System Verilog语言为基础的层次化验证平台中.结果表明,与全随机向量验证相比,该算法有效增加了功能覆盖率并使仿真时间缩短了25%左右,实现功能覆盖率的快速收敛,提高了验证效率.

With the development of large-scale integrated circuit,verification plays an increasingly important role in IC design.The functional coverage becomes a standard of IC verification.The key to improve the efficiency of verification has been a hotspot in recent years.A new functional coverage convergence technique based on genetic algorithm was proposed to improve the efficiency of verification in this article.Proportional selection operator,uniform crossover operator,and binary mutation operator were used to get the excellent verification vectors.Those genetic operators were obtained by calculating the probability distribution.An image processing chip named Turbo was selected as the verification model,and the genetic algorithm was embedded in the hierarchical testing platform built in System Verilog.Comparing with the whole random test,the proposed method reduces the generating probability of identical vectors and shortens the simulation time by 25%.In addition,the genetic algorithm improves the verification efficiency.