Regression Test Suite Generation And Minimization Research Based On UVM Verification Platform

In recent years,with the integration of integrated circuit is becoming higher and higher,the structure and function of the circuit is becoming more and more complex.In the chip development process,How to do the efficient and complete verification of the chip has become the key to limit the quality of the chip,development cost and design cycle.As an important means of chip verification,regression testing will be frequently implemented to ensure that the chip changes will not affect the original function of the chip and cause some new errors in the chip design process.However,the huge test case suite makes the cost of regression testing become very expensive.At present,how to optimize the test case suite under the premise of ensuring the completeness of the test is one of the hot issues.The mainstream test case suite optimization technology includes test suite minimization,test suite selection and test suite prioritization.This paper focus on the regression test model to discuss how to generate and minimize the test case suite,reduce the cost of regression testing,and improve the test efficiency.The mainly aspects of research in this paper are as follow:Firstly,for the generation of regression test case suite,this paper designs a regression test platform based on the UVM verification method,and automatically generates a suite of test stimuli with functional coverage driven as a regression test case,which can reduce the process of writing test cases by people.The generated test case suite will be used as the initial test case suite to do test suite minimization experiment.Secondly,for the minimization of test case suite,this paper combines genetic algorithm and greedy algorithm to propose a hybrid genetic algorithm to solve the problem of test case minimization.Genetic operators of selection,crossover and mutation are improved,which effectively improve the algorithm's global search ability.Meanwhile,the greedy algorithm to deal with feasible solutions and infeasible solutions is used,which can improve the local search ability of genetic algorithm.Finally,the standard genetic algorithm and the hybrid genetic algorithm are applied to the test case minimization problem.The experimental results show that the hybrid genetic algorithm can obtain better reduction effect and faster convergence speed under the premise of ensuring the completeness of the test.