The increasingly complex architecture of parallel computing systems makes profiling and performance measurements for high performance computing (HPC) applications in all fields of scientific computing more and more important. Nonetheless, only few researchers use the available tools and applications to optimize their applications.
In the scope of this project, the importance and significance of profiling tools shall be made visible to HPC users through automatic performance measurements. The results of these measurements will be presented to the users after each run of their application in text form as well as in a graphical manner. In addition, the user will receive tips and suggestions for the further code-optimizations on the basis of the performance measurements.
The GWDG is responsible for the project management and the dissemination of information and results obtained throughout this project.
In collaboration with other regional computing centers, performance metrics will be analyzed with respect to their significance and the acquirement options. On this basis, an interface for the collection and storage of the metrics will be developed. In a next step, the analysis and representation of the metrics will be implemented and optimization suggestions for the user will be devised. A detailed documentation and best practices will be provided and several workshops for the end users will be organized together with the project partners.