Research On Test Case Suite Reduction Technology Based On UVM

With the development of various new technologies such as artificial intelligence,5G communication,image processing,etc.,the requiments for the underlying chip technology in each application scenario are becoming higher and higher,and the circuit design is becoming more and more complicated.The speed and cost of chip development are key factors in the project.In the entire chip project,regression testing is a work that consumes a lot of resources.The test case suite will be repeatedly executed to ensure the correctness of the design,which will cause a large test operation cost.Optimizing the test case suite to improve efficiency on the basis of ensuring test completeness is one of the current research hotspots.Among them,the mainstream technologies include test case reduction,test case sequencing,and test case selection.This paper studies the test case reduction technology.The main work is as follows:Based on the actual project,this paper designs and implements a UVM-based simulation verification platform.In order to ensure that all coverpoints are covered,the test case plan is developed,and then the test cases are implemented in the verification language to form a complete set of test cases for subsequent experiments.This paper analyzes and investigates the reduction optimization technology of the test case suite and combines its characteristics to model the optimization problem as a set cover model and proposes a Memetic algorithm for solving the problem.In the algorithm,a global search strategy with a mixture of genetic algorithms and differential algorithms is designed to enhance the global search ability,and each operator is improved.A suitable local search strategy is designed based on the tabu search algorithm to avoid falling into local convergence.Then the fitness function is designed to evaluate individuals and modify the infeasible solution of the initialized population.Then simulate the test case suite to ensure that the coverage reaches the target,then collect and form a unified coverage library.Then,the standard genetic algorithm and the Memetic algorithm proposed in this paper are used to reduce the test case suite,and the results of each algorithm are compared to analyze the data evaluation performance.The reduced test case obtained by the algorithm proposed in this paper is better,time of simulation is less and the convergence speed is faster.