Specification of periscope tuning framework plugins

Performance analysis and tuning are important steps during the development of parallel software. There are performance analysis tools that help developers understand application performance. However, these tools do not give any recommendations on how to tune the code to obtain an optimized version. This paper proposes a methodology for automated analysis and tuning based on experimental space searches. The European research project AutoTune will extend Periscope, an automatic online and distributed performance analysis tool developed by Technische Universität München, with automatic online tuning plugins. Each plugin provides knowledge for performance or energy efficiency tuning. The resulting Periscope Tuning Framework (PTF) is able to tune serial and parallel codes for homogeneous and heterogeneous target hardware. The output of the framework are tuning recommendations that can be integrated into the production version of the code. The research results of AutoTune will be integrated into a commercial development environment of a European software vendor and validated with real-world applications.