Synthesis of Execution Plans for the QVT Core Language

Model transformation languages (MTLs) are important for Model Driven Engineering as they allow the automation of the engineering design process of hardware and software products, in particular at the preliminary and detailed design phases. However, the theories from compiler optimization have not been reused substantively in the development of MTLs. This makes the challenges associated with the implementation of declarative MTLs harder to overcome, in particular with respect to the synthesis of the execution plan (a representation of the control component of the transformation
algorithm). The QVT Core MTL is a declarative language, part of a set of standards proposed by the Object Management Group® in order to support the adoption of Model Driven Engineering (MDE). This research presents how instruction scheduling theories can be used for the synthesis of execution plans, in particular for the QVT Core language. The main contributions are a novel approach for performing data dependence analysis on the QVT Core language and its use for the synthesis of execution plans, and the application of metaheuristics to solve the scheduling problem inherent to the synthesis of execution plans. The research demonstrated the feasibility of applying compiler optimization techniques in the design of MTLs and provides a methodology that can be used to construct effi cient execution plans that result in correct transformations. The performance gains and correctness will help the widespread use of the QVT Core language and encourage the adoption of compiler optimization techniques in the implementation of other MTLs.