Section: New Results

Discrete control and reactive language support

Participants : Gwenaël Delaval, Eric Rutten, Stéphane Mocanu, Alia Hajjar, Abdoul-Razak Hassimi Harouna.

Concerning language support, we have designed and implemented BZR, a mixed imperative/declarative programming language: declarative contracts are enforced upon imperatively described behaviors (see 6.1 ). The semantics of the language uses the notion of Discrete Controller Synthesis (DCS) [5] . This work is done in close cooperation with the Inria team Sumo at Inria Rennes (H. Marchand).

New results concern the master internship of Alia Hajjar, co-directed by Gwenaël Delaval an Stéphane Mocanu, on the subject of Application of control of reactive environments and probabilistic models on Transactional Memory. Multiprocessor environments which use concurrent programs and data structures showed the need of techniques to organize the usage of the shared structures, to reduce the unpredicted delay and reduce the contention between concurrent processors. Transactional Memory (TM) is a programming model that eases development of concurrent applications. Concurrent programming causes conflicts and TM is a way to resolve these conflicts with the transaction paradigm. To control conflict, techniques are provided to optimize (identify the best) degree of parallelism. In this framework, the aim is to control the TM system by adapting the degree of parallelism in order to maximize the throughput, i.e., number of committed transactions per time. The main objective is to minimize the execution time of a parallel application, thus maximize the throughput. During this master's thesis, the behavior of a multithreaded TM environment has been modeled as a stochastic discrete event system. The Heptagon/BZR language has then been used to implement this model for simulation, and evaluation of control strategies.

Ongoing work concerns aspects of compilation and debugging and exploring the notion of adaptive discrete control, which is yet an open question in discrete control in contrast to the well-known adaptive continuous control.

Another activity related to discrete control is or work with Leiden University and CWI (N. Khakpour, now at Linnaeus U., and F. Arbab) on enforcing correctness of the behavior of an adaptive software system during dynamic adaptation is an important challenge along the way to realize correct adaptive systems [11] .