基于遗传算法的随机测试生成技术探究

GENETIC ALGORITHM BASED RANDOM TEST GENERATION: A CASE STUDY

随着集成电路设计复杂度不断提高,功能验证的挑战也不断增大。为了加快验证进程提高覆盖率,提出了一种新的基于遗传算法(Genetic Algorithm,GA)的随机测试生成方法。该方法基于一种二进制和十进制数混合编码的双层编码模式,并使用了权值可自动调控的功能覆盖点来计算个体适应度值,并将模拟过程中的覆盖率报告自动反馈回给随机测试生成的约束产生。该方法已应用于中国科学院微电子研究所自主研发的IME-Diamond数字信号处理器RTL模型的模块功能验证。实验结果表明,该方法有效提高了验证效率。

Abstract As the complexity of the hardware design increases,so does the challenge of functional verification.Random test generation technology is now commonly used in simulation based verification.In order to speed up the verification process,coverage directed test generation(CDG) has been proposed to lead to better coverage rate.This paper presents a new genetic algorithm based approach to close the feedback loop between coverage report and directives for random test generation.A double-layer encoding mode,which is binary and decimal mixed,is adopted.The functional coverage points,which weights are adjustable automatically,are used to calculate individual fitness.The experiment result shows that the algorithm promotes the efficiency of verification on IME-Diamond DSP processor in RTL level.