Improving software performance is a vital prerequisite for its usability. Accessibility to powerful peta- and in near future exa-scale system offers great advantages to codes that can utilize them efficiently. Various specialized processors such as GPGPUs and FPGAs can offer a tremendous value as cheaper or more power efficient accelerators.
In this section, the Competence Center presents an overview of the specifics of HPC computing, GRID computing and some less popular accelerators and platforms. Results from the analysis and optimization work within the ScalaLife project can be a useful guide for other developers, while users of the software packages can find the benchmarking and validation results useful on understanding what to expect from the different hardware platforms.
Efficient utilization of the upcoming exa-scale system would most likely require development of ensemble computing techniques. They are mainly concerned with utilization of inherent parallelism on the problem level rather than working on the algorithms that are implemented in the application. In ScalaLife we have successfully used the COMPSs programming model for efficient parallelization of DISCRETE.