The EU funded “Performance Optimisation and Productivity (POP) Centre of Excellence (CoE) in Computing Applications” currently provides various services to help improve performance of parallel software, including software written using MPI and OpenMP and for GPUs.
Often parallel software suffers from unexpected performance problems, for example because of communication and synchronisation issues. These issues can lead to long run times or increased hardware expense, and may place fundamental limits on what problems can be solved.
Solving parallel performance problems typically requires investing time to learn sophisticated performance analysis tools and to become expert in interpreting complex runtime data. The POP CoE has been funded to provide access to the necessary tools, skills and expertise.
These services are free of charge to organisations within the EU, and so are ideal for academic researchers who need help understanding why parallel software is not performing well, and to assess how code could be improved.