Abstract

We introduce a SAT based auto338m abstraction refinement framework for model checking systems with several thomGG4 state variables in the com o influenceo f the specificatio8 The abstractmo del iscoK060mEN8 by designating a large numbero f state variables as invisible. In co trast to previoN wo rk where invisible variables were treated as free inputs we describe a co06NGmEG7430m mo0 advantageo3 approF h in which the abstract transitio relatio isappro ximated by pre-89889L6728 invisible variables during imageco8087FmEG0 The abstract co4 terexamplesorexamp fro mo del-checking the abstract mo del are symbo lically simulatedo the coG0K8K system using a state-oGNK7Kmo SAT checker. Ifno co43FK3 co4 terexample isfo640 a subseto f the invisible variables is reintro duced into the systemand thepro cess is repeated. The main co tributio o f this paper are two new algo37FmE fo identifying the relevant variablesto be reintro duced. Thesealgo78NNm mogo7 the SAT checking phase inom4F to analyze the impacto individual variables. Ourmetho d is co48NFF fo safetypro erties (AG p) in the sense that -- perfoN06G0 permitting -- a pro erty is either verifiedo dispro ved by aco4GKKm co4 terexample. Experimental results are givento demoGGmE40 the power of our method on real-world designs.

Citations