Multiprocessors and multicore processors are now ubiquitous, but programming these systems, to deliver high-performance and reliable systems, is very challenging. Shared-memory is the programming abstraction exported by the hardware, and most multi-threaded programs communicate through memory shared between the threads. Traditionally concurrent execution was viewed as simply an interleaving of the steps from the threads participating in the computation. Thus if we started in an initial state in which all variables are zero, and one thread executes: x = 1; r1 = y; while another executes y = 1; r2 = x; either the assignment to x or the assignment to y must be executed first, and either r1 or r2 must have a value of one when the execution completes. However, it has proven impractical to guarantee such a restrictive memory behaviour, and all realistic programming language implementations supporting true concurrency allow both r1 and r2 to remain zero in the above example. There are two reasons for this: - for efficiency reasons, compilers may reorder memory operations if that doesn't violate intra-thread dependencies; - the hardware may reorder memory operations based on similar constraints. The forthcoming revision of the C++ standard (the C standard will be updated accordingly to preserve compatibility) specifies all the constraints that the possible outcomes of a parallel program must respect. This requires special architecture-dependent support in C and C++ compilers. The goal of this grant proposal is to investigate the formal verification of realistic compilers for concurrent dialects of the mainstream languages C and C++. We will target both the x86 and Power/ARM architectures, which require radically different compilation strategies and proof methods, focussing initially on sample compilation schemes and then lifting these results to fully-fledged compilers. In addition we will design and prove correct novel compile-time optimisations for these languages and compilers.

Existing Warehouse Management Systems (WMS) provide advanced features to manage the movement of items within the warehouse, but fail to comply with the increasing demand on more numerical handling, in particular for achieving a flexible supply chain model capable of handling personalised orders in a cost-effective and eco-efficient manner. Generally, WMS are lacking optimisation functionalities and advanced packing tools for determining how to pack items on a pallet, how many cartons are needed to pack customer items, how to pack pallets in a truck according to stability constraints and customers to visit, or at a larger scale, how to redesign a storage area, an assembly line, etc. The vision of the Net-WMS-2 project is that these hard combinatorial optimisation functionalities can be addressed in a new generation of WMS with advanced software technologies combining rule-based knowledge representation and constraint programming optimisation. This project builds on the former European FP6 Net-WMS Strep project that has shown that constraint-based optimisation techniques can considerably improve industrial practice for box packing problems, while identifying hard instances that cannot be solved optimally, especially in industrial 3D packing problems with rotations, the needs for dealing with more complex shapes (e.g. wheels, silencers) involving continuous values. The aim of the Net-WMS-2 follow-up project is to solve these remaining problems by focusing on them with a restricted consortium composed of the three key partners on these particular topics. To this end, we expect to advance the state-of-art in - constraint solving for hybrid discrete-continuous geometrical constraints, - constraint propagation with polymorphic shapes, - search strategies for packing problems with discrete rotations, - knowledge representation with rule-based modelling languages, - integration of design and optimisation technologies in WMS. Based on these contributions, we will develop innovative prototype tools for supply chain decision making, including: - a generic pallet/container loading optimiser - a prototype solving a real-world industrial pallet loading problem - a packing designer for complex shapes - a prototype solving a real-world packing problem with complex shapes The academic partners will make most of the technology available to the scientific and technological communities through their integration in the open-source software they develop, respectively in the CHOCO/IBEX constraint programming system developed by EMN, and in the Rules2CP modelling language developed by INRIA. The SME KLS OPTIM will integrate these developments in the existing KLS optimisation suite which is itself largely based on Net-WMS technology. In this way, we hope, on the one hand, to make fundamental contributions in the emerging field of hybrid continuous-discrete geometrical constraint programming, and on the other hand, to apply these novel optimisation techniques to real-life packing problems in the industry, with a flexible supply chain model capable of handling personalised orders in a cost-effective and eco-efficient manner.

System identification is about the theory and the practice for building mathematical models of dynamic systems from experimental data. This topic is common in many areas of sciences and technologies, though the term of “system identification” is usually used in the field of automatic control, where it is studied with the particularity of control systems. As a matter of fact, system identification plays an important role in the design of modern control systems, as most efficient controllers are model-based. The main idea of block-oriented nonlinear system identification is to model a complex system with interconnected simple blocks. Such models can cover a large number of industrial applications, and are yet simple enough for theoretic studies. It is therefore a good trade-off between the studies on general nonlinear systems producing few practically useful results, and those on specific nonlinear systems with a limited application scope. The objectives of the proposed project are to extend block-oriented nonlinear models with hysteresis blocks and bilinear blocks, and to relax some traditional restrictions on nonlinearity structures and on experimental conditions. The two extensions with hysteresis blocks and bilinear blocks have been motivated by their importance in process control. Through these extensions, it is expected to considerably increase the applicability of block-oriented nonlinear system identification to industrial systems, while contributing significantly to the progress of the researches on nonlinear system identification. Two case studies are planned in this project, namely, a control valve with stiction, and a fuel cell system, both with direct industrial backgrounds. These two case studies will not only serve as laboratory validations of the produced results, but also demonstrate their feasibility in industrial applications. The outcome of this project is expected to cover more industrial applications with the extended block-oriented models and to better adapt the identification methods to industrial application environments. By means of increasing the efficiency and the reliability of model-based control and monitoring systems, the benefits of this project may impact industrial production quality, energy and raw material saving, human and equipment safety, and environmental protection.

Logic and algorithms are two important research areas in computer science. Traditionally, logic has been one of the important cornerstones of mathematics in Europe, while Chinese scholars have been more adept in the research on algorithms. The famous "Nine Chapters on Algorithms" in ancient China and the contemporary mathematician Professor Wu Wenjun’s mechanical proof theory are precisely the incarnation of traditional Chinese philosophy. The dialogs between logics and algorithms have been a constant source of progress for both disciplines, which, in turn, stimulates the development of computer science. The dialogs are reflected in two aspects: one aspect focuses on the languages used to express proofs and algorithms. The algorithmic interpretation of proofs permits the use of algorithmic languages as proof languages, i.e. to import languages from computer science to logic, and conversely to equip proof languages with an operational semantics and to give computer science new algorithmic languages, by importing them from logic. Another aspect of this dialog is the challenge to incorporate concurrent programs. This duality is reflected by the distinction between lambda-calculus and pi-calculus. Our project extends and reformulates both calculi and proposes to design new programming languages and proof systems, based on a study of their relation to logic. The research of this project will focus on new emerging directions in this dialogs between logics and algorithms : namely, a new logically inspired computational method : the scheme calculus, a new way to express theories in a computationally inspired way in clausal polarized Deduction modulo, a new extension of tree automata with internal communication and parallel composition, that extends also the language CCS: CCTS, and a new calculus to solve distributed constraint optimization problems and other similar problems (e.g., multi-core scheduling) from the perspective of co-induction. The members of the project have been active in the emergence of these new calculi. The project has not only theoretical significance but also practical relevance.