Given the increasing data volumes and increasing complexity of algorithms, there is an ever-growing need for computational power. To render the computational work feasible our research encompasses efficient implementation strategies, parallel computing and performance modeling/analysis.

This includes the following topics:

• Personal high-performance computing: we study the use of accelerators such as GPUs, FPGAs or upcoming architectures such as Intel's MIC to turn a standalone computer into a 'number cruncher'.
• Improved strategies for common communication operations on large-scale systems.
• Performance analysis and modeling for feasibility studies and performance bottleneck detection.
• Data mining on empirical data from the performance analysis of algorithms and architectures.
• Memory optimizations.
• The knowhow for lowering the threshold to exploit the processing power offered by modern accelerators.