Model-Based Verification and Estimation Framework for Dynamically Partially Reconfigurable Systems

Abstract

Compared to the verification of a conventional embedded system, that of a dynamically partially reconfigurable system is more complex and difficult because the requirements for partial reconfiguration changes with the time and environment. Thus, a complete and efficient verification mechanism is urgently required. Different from the non-exhaustive simulation-based verification used in most UML-based design flow for dynamically partially reconfigurable systems, we propose a model-based verification and estimation framework (MOVE) that covers the function-oriented platform independent verification and the physical-aware platform specific verification based on the model-driven architecture (MDA) design. Userspecified UML models are first exhaustively verified using model checking and then integrated into a UMLbased hardware/software co-design platform (UCoP) for physical-aware verification and estimation at a high abstraction level. Our experiments also demonstrate that the state-space-explosion problem does not occur when model checking reconfigurable systems using MOVE. Furthermore, we can have more accurate time measurements using UCoP which can be used to validate system correctness and performance at the system level compared to the existing synthesis-based and lower-bound estimation methods.