A Brief Comparison Sample Clauses

A Brief Comparison. Comparing the di erent meta-models, we nd many similarities. On the sur- face, however, there may seem to be more di erences than what is really the case. The di erences stem from various sources. One obvious reason is di erent naming of similar constructs. Another reason is that sometimes, a property of a construct in one model is made explicit as a separate con- struct in another. Also, particular constructs may be completely lacking in one model, but exist in another. Finally, there is naturally the possibility that errors have been made in the metamodeling, since we in most cases have transformed a textual description into the graphical models. We make a simple comparison by listing corresponding constructs in the di erent models, shown in Fig. 2.42. Doing this, we also highlight the three former reasons for di erences between models. We will use the ontological model as a basis for comparison, since it contains a few, basic constructs, and since it has already been used for the purpose of analyzing CMLs. The constructs Environment, System, and Value&Time are omitted, since these are not found in the other models. From the table, we see that Xxxx and Xxxxxx DRAFT January 2, 2000

A Brief Comparison. We briefly compare techniques for analyzing pointer programs based, on the one hand, on separation Logic and shape Analysis [115] and, on the other hand, on HRGs. As mentioned earlier, separation logic is an extension of Hoare logic with inductive defini- tions to describe the shape of heap parts. It is classically employed in the form of Hoare-style verification using annotations of programs where decidability of entailment is essential. Whereas this problem is undecidable in general, decidable sub-logics for lists and trees have been de- veloped in [13, 14]. There are several separation-logic tools such as Space Invader [125, 27] (supporting linear data structures), Xxxxxxxxx [14] (aiming at (doubly) linked lists and trees), jStar [43], and Verifast [76] (allowing general user defined predicates and data structures). The main advantage of these tools is their scalability [125]. The HRG-method is based on exhaustive (abstract) state space exploration and can support a rich set of user-defined data structures. As pointed out by Xxxxx and Plump in [44], there exists a strong correspondence between inductive definitions in separation logic and the nonterminals that are used in the abstract heap representations based on graphs. This also means it would be interesting to investigate if this relationship can be exploited e.g., to allow for usage of user- defined data structures given by HRGs in separation logic contexts while preserving decidability of entailment. Shape analysis is based upon three-valued logic, where the value “don’t care” is used to deal with loss of information caused by abstraction. Heaps are described by shape graphs, where nodes, indistinguishable by properties expressed as predicated in the logic, are summarized. Typical predicates are shape properties like reachability, cycle membership, etc. Most of these are derivable from the heap graph representation [38] and are implicitly provided in the HRG state space exploration. Whereas in shape analysis all nodes satisfying the same predicates are summarized, in the HRG approach nodes that constitute a well-defined (fragment of a) data structure are collapsed.