Self-Adaptable Virtual Machines

In the past few years, researchers in software language engineering tried to reduce the costs of programming language creation. This comes with different approaches like improvement of language workbenches usability and language modularity and composition. With those approaches, the design of a new language became an affordable solution to address domain-specific problems. However, such a generic approach hamper the application of fine-grained optimizations within the resulting language, together with energy-demanding runtimes.

We propose to explore the definition of a new design pattern, and the corresponding software architecture, to develop an interpreter in the form of a dynamically adaptable system. The architecture should reify an explicit feedback loop, with a possible reconfiguration for the sake of performances, energy consumption, or a trade-off between them. The candidate will review the literature about software performance benchmarking, language design and implementation (esp. metamodeling techniques), dynamically adaptable systems engineering (e.g., MAPE-K loop), and program transformation tools for static and dynamic analysis (e.g. SPOON). Based on the literature review, the candidate will propose a new design pattern to enhanced the implementation of a virtual machine with a feedback loop and will evaluate it on various domain-specific languages.

The PhD will take place in the DiverSE team at IRISA, and in collaboration with KTH, Sweden (Group of Prof. Monperrus, CASTOR center). The candidate will join a research project already initiated 1 year ago and will be involved in a group of PhD students and Professors working on the topic.