Effects of property ordering in an incremental formal modeling methodology

Abstract

In this paper, we analyze the effect of ordering linear time properties while using the Extreme Formal Modeling (XFM) methodology in building "prescriptive formal models" (PFM). PFMs are formal models built incrementally by adding user stories and are used as specification golden models. In our methodology, the user stories are captured in Linear Time Temporal Logic (LTL). A more expressive logic or formalism could be used for describing the user stories as well. During incremental model building, the PFMs often blow up in size in terms of the state space, and the main tenet of XFM being regressive model checking, blown up models often make it impossible to carry out the XFM methodology. Here, we propose property ordering hueristics to circumvent this problem. We compare these hueristics with: (i) no specific ordering of user stories (standard approach), (ii) sorting of the user stories based on a weighting scheme (property based sorting), and (Hi) predicate based sorting of user stories based on an eliminative scheme (predicate based sorting). We show that the predicate based sorting scheme is the most effective way to carry-out XFM model building. We illustrate the schemes and the comparison by modeling a monitor for the ISA bus and for the arbitration phase of Pentium Pro processor's bus using the Cadence SMV. We also provide an algorithm for the predicate based sorting that yields the best control on the increments in model size.