Test Generation (MuProd)     


Ronen Levy photo

Test Generation (MuProd) - overview

Test generation for simulation has developed greatly in the past decade in the hardware verification domain. This was due to the high cost of hardware bugs and also due to the use of abstract formal models as part of the HW development process which enabled the use of sophisticated technology. The advance of test generation was carried out together with corresponding advancements in related simulation-based verification techniques - like monitors, checking, coverage collection and analysis, and constraint-modeling and satisfaction, and various simulation and hardware emulation techniques. The IBM research labs at Haifa have played a leading role in this development and have advanced the dynamic verification methodology used to verify IBM's hardware systems. This is based on expert systems for the various types of tested systems. The expert system is based on an external anthology that models the tested system and testing knowledge describing how best to test the system. The expert system is driven by a verification engineer that describes the desired test by using a dedicated "test-template" language tuned for his specific testing domain. The methodology is complemented with appropriate checking techniques to check for possible failures during the test, and with with coverage analysis capabilities that help to assess the progress of the testing process and to modify the test-templates if needed. Our goal now is to duplicate the success of this sophisticated dynamic verification methodology to new fields of systems-engineering -specifically for the domain of testing production processes, and for the domain of testing in-car networks. In both these cases we will adapt an IBM expert system tool for test generation of hardware systems (X-Gen). The work on production process testing will is carried out in the context of the EU project - MuProD which deals with introducing in-line testing techniques to production processes. IBM's role will be to develop a model-based test-generation expert system for simulation of production systems at the high system level ("production-chain" level). In the automotive context we will develop an expert system to test in-car networks - which in principle are similar to the types of hardware systems tested at IBM.