New Methodology for Complete Properties Extraction from Simulation Traces Guided with Static Analysis

Abstract

This paper introduces a new methodology for digital design properties extraction from simulation traces. A new Breadth-First Decision Tree (BF-DT) mining algorithm is proposed for complete properties extraction from simulation traces. The new mining engine supports both bit-level and word-level values of different design variables. Static analysis technique is implemented to extract all data dependencies between the digital design variables. New traces regeneration algorithm is proposed to obtain reduced traces for more efficient and easier mining. The mining engine is guided with these data dependencies to extract complete design properties relating target variables desired to extract properties for and their cone of interest feature variables. The contributed mining technique has been tested for different designs with different sizes. The design properties generated from the mining engine completely match with all design properties covered in the input simulation traces. Moreover, the generated properties are at the highest possible level of abstraction leading to the best coverage for the input data space. The simulation results show that the proposed methodology has superior efficiency in extracting both bit-level and word-level complete assertions of digital design in both superior quality and feasible time.