Modified Adaptive Differential Evolution Algorithm for Test Scheduling of Multi-Core SOC Based on DVS and MVI

摘要

As a popular kind of system chip, multi-core SOC based on DVS and MVI has the advantages of low power consumption and high performance. However, the test technology is still in the exploration stage, being the first problem restricting the development of multi-core SOC. In this paper, a new modified differential evolution algorithm (JADE-MaS) is applied to system-level test optimization. JADE-MaS proposes a multiangle searching strategy and obtains good effect on enhancing population diversity. The mathematical model of TAM resource division and test scheduling include four decision variables: the test bus width, the location of IP cores, the start and end times of the test. We encode the first two variables as population individuals and use JADE-MaS to seek the optimal individual, then use a priority scheduling mechanism based on heuristic algorithm to distribute the test tasks to evaluate the individual fitness. The remaining variables can also be solved in this process. After being decoded, the individual with the shortest test time is the solution of system-level test scheduling problem. Series of experiments on ITC'02 SOC benchmarks show that JADE-MaS has searched the better scheme comparing with the GA and the PSO algorithm and effectively shortens the system-level test time.