Recent developments in HPC for scientific applications
Numerical simulations have become a key method for research in different academic disciplines, especially in natural sciences. To provide the required resources within a given cost and power envelope the architectures of HPC machines are becoming more complex and heterogeneous. Therefore, the issue of usability of new hardware architectures becomes more critical. In this talk we review recent developments in HPC architectures, confront them with application requirements and provide some outlook on future trends.