My research interests are twofold:
My primary concerns were the definition of the language constructs and associate concepts
for conciliating program reusability and efficiency – for instance, the alignment concept for expressing data locality in some data parallel languages
and the notion of directives to the compiler in languages such as HPF.
In some sense, such constructs and concepts are means for the programmer to express in an abstract and elegant way the code transformations that compilers should apply to the code. Therefore, an aspect of my research is the study of code transformations and especially their formal basis. Recent research interests focus on the use of the grid which brings with it some new challenges due to its heterogeneity and dynamicity. I am investigating some “generic” code transformations in order to make codes more flexible and parameterized by a grid state. These transformations are originally designed for and validated on a real application in Plasma Physics which is a representative candidate for the grid.
Parallel programming – The data parallel programming model -
Semantics of programming languages – Program refinement – Code transformations – Directives to the compiler -Grid computing -
Since its creation, I am a member of the INRIA project CALVI. This project is devoted to the mathematical and numerical study as well as to the visualization of different problems arising mostly from plasma physics and beam physics. One major difficulty in these applications is the existence of multiple scales and the large size. Therefore an important effort regarding modeling and development of simulation and visualization methods is needed. The project involves applied mathematicians and computer scientists specialized in parallel computing and visualization with strong links to plasma physicists.
Adaptive numerical methods – Vlasov equation – Heterogeneous architectures -