Computational Analytics

This Lab has an upper limit on the number of participants. If you want to participate, please write an informal application e-mail to Gayatri Manda (kmanda@uni-bonn.de) until April 13, 2026.

In this course, participants will gain valuable hands-on experience by working directly with a real-world scientific application, executed on a state-of-the-art supercomputer Marvin in University of Bonn. The course begins with learning how to profile and instrument code, giving you the understanding of the tools to deeply analyze performance behavior. You'll conduct scaling studies to understand how applications perform under various parallel configurations, applying core concepts to interpret and explain results. These investigations will help you identify performance bottlenecks and limitations in the current implementation.

As the course progresses, you'll explore a range of optimization strategies to improve the performance of the application. This includes experimenting with thread and process placement, memory access patterns, and load balancing techniques, among others. You’ll also apply and refine parallel programming best practices. The emphasis is not only on making the code faster but also on understanding why certain changes work and how to make your optimizations portable and scalable across systems. This experience provides a solid foundation in high-performance computing (HPC) and prepares you for real challenges in research.

Throughout the lab, participants will be given a series of short, focused exercises designed to build their skills incrementally in performance analysis and optimization. By the end of the course, you are required to submit a final report that presents the analysis and improvements so far. The lab ends with a final presentation on some specific topic after which you will also be asked some questions concerning the content of the lab.