My research focuses on the development of HPC libraries for different physical problems.
It includes the creation of new parallelization strategies using fork-join paradigm (OpenMP), message passing interface (MPI) or runtime systems (StarPU), but also all the modules to have an industrial quality package.
Therefore, I develop low-level kernels for CPUs (SSE/AVX/ASM) and GPUs (CUDA/OpenCL), and various complex algorithms.
I usually work in C++ and make the effort to have a clean code and a maintainable design.
For five years, I have worked on the Fast Multipole Method algorithm and the implementation of a time-domain BEM for the wave equation.
I am now involved in the development of a protein simulation application.