寄存器传榆级(RTL)描述是目前应用最广泛的电路设计描述形式.在时序电路的RTL激励生成中,基于模拟的方法避免了帧扩展法庞大的搜索空间,但采用该方法常存在向量过多.质量不高等问题.本文充分考虑影响算法效率的各种因素,在此基础上,提出一种基于混合遗传算法的激励生成方法.该方法结合多种覆盖评估准则与故障模型作为激励生成目标,同时采用动态参数设置,实现全局-局域混合搜索机制.实验结果显示该方法是有效的.
The Register Transfer Level (RTL) behavioral descriptions are widely used in IC designs. In pattern generation for sequential circuits at RTL, simulation-based methods avoid the large search space used in time-frame expansion methods,but the quality of patterns can't be guaranteed often. This paper proposes a pattern generation method based on hybrid genetic algorithms. Firstly a global-local hybrid search mechanism is adopted in this system; secondly, many coverage metrics and RTL fault model are taken into account; thirdly, dynamic parameter settings are used to enhance the efficiency of the method. At last the experimental results show the effectiveness of the proposed method.