TAGTransformation and Adaptations for the Grid

The TAG project aims at improving the utilization of parallel applications on computational grids. Our long-term objectives are mainly related to programming models for grids, which implies many other issues, such as load-balancing in heterogeneous environment, middleware fault-tolerance, etc.


PCOProgram Compilation & Optimization

The aim of this project is to provide automatic methods for performance optimization of programs. At this time, our activities focus mainly on iterative program structures (nested loops), accessing array elements from affine reference functions. The considered loops are characterized by affine loop bounds expressed on the loop indices and on some parameters. This model allows to address many matrix computation algorithms in scientific computing, as well as multimedia applications for embedded systems like encoding/decoding and numerical filters. The symbolic characteristics of our approach allows to construct adaptive methods for different memory architectures and problem sizes



In many scientific domains, simulation is a fruitful research topic beside theoretical and experimental approaches. In this field, Image and High performance Computing are very promising techniques. Our Group has worked on various application domains.

this Reconstruction of bounding surfaces in 3D images [More]
this Parallel algorithms for Image registration for medical applications [More]
this Parallel seismic Ray Tracing in a global Earth model [More]
this Optimization tools for scientific applications [More]
this Data distribution for a parallel Vlasov solver in Plasma Physics [More]
PolyLibA library for manipulating parameterized polyhedra

The Polyhedral Library (PolyLib for short) is an independent C library manipulating objects made up of unions of polyhedra of any dimension. It was first developed by Doran Wilde at IRISA, in Rennes, France, in connection with the ALPHA project. Many people contributed to the developement of the library in the past years. Current developments are devoted to parametric computations. In particular, the latest release, available here, includes parametric vertices computation and Ehrhart polynomials computation.


PeiParallel Program Semantics

The PEI equational language gives the programmer the ability to define relations between variables called data fields, the whole set of equations being the program. One set operator and three functions allow to write programs in a data parallel way, and then transform the program through a refinement process until a semantically equivalent version is acceptable for execution on a parallel architecture.