Implicit Incremental Model Analyses and Transformations

In many engineering disciplines, abstract models are used to describe systems on a high level of abstraction. On this abstract level, it is often easier to gain insights about that system that is being described. When models of a system change – for example because the system itself has changed – any analyses based on these models have to be invalidated and thus have to be reevaluated again in order for the results to stay meaningful. In many cases, the time to get updated analysis results is critical. However, as most often only small parts of the model change, large parts of this reevaluation could be saved by using previous results but such an incremental execution is barely done in practice as it is non-trivial and error-prone. The approach of implicit incrementalization o ers a solution by deriving an incremental evaluation strategy implicitly from a batch speci cation of the analysis. This works by deducing a dynamic dependency graph that allows to only reevaluate those parts of an analysis that are a ected by a given model change. Thus advantages of an incremental execution can be gained without changes to the code that would potentially degrade its understandability. However, current approaches to implicit incremental computation only support narrow classes of analysis, are restricted to an incremental derivation at instruction level or require an explicit state management. In addition, changes are only propagated sequentially, meanwhile modern multi-core architectures would allow parallel change propagation. Even with such improvements, it is unclear whether incremental execution in fact brings advantages as changes may easily cause butter y e ects, making a reuse of previous analysis results pointless (i.e. ine cient). This thesis deals with the problems of implicit incremental model analyses by proposing multiple approaches that mostly can be combined. Further, the