分类树在分组密码IP验证中的应用与改进

Application and Improvement of Classification Tree in Block Cipher IPs Verification

分组密码IP核具有配置过程复杂、数据运算量大的特点,如何对其进行高效的验证是整个设计面临的关键问题.在随机验证中,激励生成和覆盖率模型抽象占据着尤为重要的位置.本文将分类树方法应用于分组密码IP的功能验证,并且针对其无法解决关联操作和顺序控制的缺点实施改进,主要是引入虚拟输入对激励序列进行规划,构建超长输入数据包.实验证明,采用改进的分类树指导激励生成和覆盖率模型抽象,能够生成更加精简有效的激励和完备的覆盖率模型,进而显著地提高验证的效率和完备性.

The features of block cipher IPs are complex configuration process and massive data operation. How to implement the veri- fication efficiently is the critical issue throughout the whole design process. Stimulus generation and coverage model abstraction occu- py very important position in the random verification. This paper applies the classification tree method to the functional verification process of block cipher IPs, makes improvements on the shortcomings in association operation and sequence control, and proposes virtual input to make a plan for stimulus sequences aiming at constructing the super-long packets. Experiments show that applying im- proved classification tree method to guide stimulus generation and coverage model abstraction can generate more effective stimulus and complete coverage model, and significantly improve the efficiency and completeness of verification.