The aim of this project is to develop a theoretical and computational framework capable of describing the dynamics of a multielectron atomic system exposed to strong and/or ultra-short radiation, particularly the free-electron laser produced radiation. In terms of radiation characteristics the applicability of the theory ranges from the far infrared (fraction of eV) to the soft X-ray (keV) spectral band and pulse durations from attoseconds to hundreds of femtoseconds.

In the past two decades there have been several changes in various directions in the computational infrastructure. The Beowulf-CPU-based approach, constitutes a very attractive supercomputing architecture for medium size institutional and academic organizations, since its requirements in terms of economical, hardware, software and technical support resources are the lowest possible.

In this project we apply the GPU programming model, that has appeared recently due to the availability of higher level languages, such as the OpenCL variant of C99, which are compiled to run directly on the GPU. The major advantage of the GPU architecture is the large number of cores present inside a single GPU (over a 1000 for the most powerful). Thus a desktop supercomputing environment appears feasible in the next coming years. Furthermore, a Beowulf-GPU based model approach will offer a very powerful supercomputing workstation at low cost, which can be a strong-candidate solution for medium-size institutes.

We intend to develop black-box GPU-based routines which perform three particular mathematical operations, namely the matrix-vector multiplication, linear equation solution and matrix diagonalization. Our goal is the drastic reduction of the computational cost for a number of important scientific problems.