Maximal Causal Models for Sequentially Consistent Multithreaded Systems
Șerbănuță, Traian Florin
This paper shows that it is possible to build a theoretically maximal and sound causal model for concurrent computations from a given execution trace. For an observed execution, the proposed model comprises all
consistent executions which can be derived from it using only knowledge about the execution machine. The existence of such a model is of great theoretical value. First, by comprising all feasible executions, it can be used to prove soundness of other causal models: indeed, several models underlying existing techniques are shown to be embedded into the maximal model, so all these models are sound. Second, since it is maximal, the proposed model allows for natural and causal-model-independent deﬁnitions of trace-based properties; this paper proposes maximal deﬁnitions for causal dataraces and causal atomicity. Finally, although deﬁned axiomatically, the set of traces comprised by the proposed model are shown to be eﬀectively constructed from an initial observed trace. Thus, maximal causal models are not only theoretically relevant, but they are also amenable for developing practical analysis tools.