From July 14th to 18th, 2025, the 15th ACM Europe Summer School on Programming and Tuning Massively Parallel Systems + Artificial Intelligence (PUMPS+AI) took place at UPC Campus Nord in Barcelona. This prestigious event brought together researchers, graduate students, and instructors from around the world to explore the latest advances in massively parallel computing and artificial intelligence (AI).
M. Augusto Maidana Silanes, member of the Fusion Group at the Barcelona Supercomputing Center (BSC), had the opportunity to attend this year’s edition. The school provided an exceptional environment to gain both theoretical knowledge and hands-on experience in programming for many-core processors and GPU accelerators, with a special emphasis on AI applications.
The 2025 program offered a sneak peek of the upcoming 5th edition of Programming Massively Parallel Processors, a widely recognized textbook in the field. Participants engaged in lectures and practical sessions covering topics such as:
- Large language models
- Matrix multiplication optimization
- Wavefront algorithms
- Sparse and dynamic data handling
- Dynamic parallelism and multi-GPU execution
- Debugging and profiling CUDA code
With lectures from world-leading experts such as Wen-mei Hwu (NVIDIA/University of Illinois), Juan Gómez-Luna (NVIDIA), and Izzat el Hajj (American University of Beirut), along with BSC and UPC specialists including Antonio J. Peña and Marc Jordà, the event highlighted both the theoretical foundations and real-world implementations of GPU-accelerated applications.
For BSC researchers, events like PUMPS+AI are a key opportunity to strengthen knowledge, foster collaborations, and stay at the forefront of high-performance computing research. In particular, the combination of GPU architecture expertise and the expanding role of AI is essential to tackle the complex computational challenges in nuclear fusion research, one of the most demanding frontiers in science and engineering. This synergy underscores the critical importance of massively parallel computing in driving breakthroughs for future sustainable energy solutions.