CGA: Combining cluster analysis with genetic algorithm for regression suite reduction of microprocessors

Abstract

Regression testing plays an important role in the simulation-based functional verification of microprocessors. Regression suite is maintained in the entire verification phase with an increase of the scale. However, the executing cost is always high when running the entire suite on a RTL-level simulator. Regression suite reduction (called RSR for short) is presented to reduce the executing cost of the regression suite without debasing the quality of the functional verification. For this two-objective RSR of microprocessors, we present a heuristic algorithm which mainly combines cluster analysis with genetic algorithm (called CGA for short). The experiments on some regression suites at different scales for a microprocessor have shown the efficiency and feasibility of CGA. CGA can effectively reduce about 90% of the executing cost without decreasing the functional coverage in an acceptable runtime.