Self-adjusting constrained random stimulus generation using splitting evenness evaluation and XOR constraints

Constrained random stimulus generation plays significant roles in hardware verification nowadays, and the quality of the generated stimuli is key to the efficiency of the test process. In this work, we present a linear dynamic method to guide random stimulus generation by SAT solvers. A splitting simplified Min-Distance-Sum evaluation method and an XOR sampling strategy are integrated in the self-adjusting random stimulus generation framework. The evenness of the split groups is evaluated to find out some uneven parts. Then, random partial solutions for the uneven parts and random XOR constraints for the other inputs are added into constraints to get better distributed stimuli. Experimental results show that our method can evaluate the evenness as well as more complex formulae for stimulus generation, and also confirm that the self-adjusting method can improve the fault coverage ratio by more than 17% averagely with the same number of stimuli.