PRACE Autumn School familiarised international students and postdoctoral researchers with modern GPU-accelerated supercomputers

Autumn School participants on their visit to the LUMI data center. Image: Fredrik Robertsén, CSC

The PRACE Autumn School: Harnessing the EuroHPC Flagship Supercomputers in Vuokatti, Finland (11-15 October 2021) brought together undergraduate and graduate students and postdoctoral researchers in different disciplines of science using scientific computing. The five-day school, organised by CSC – IT Center for Science, was filled with lectures, hacking, leisure activities, and having a good time together in Northern Finland. The PRACE Autumn school was the first face-to-face event after a long time. Luckily the covid situation in the Kainuu area was rather calm during the school. In any case, the course considered the measures of preventing COVID-19 spread.

One of the highlights of the school was the visit to CSC’s data center in Kajaani, the home of one of the most powerful supercomputers in the world: LUMI. It was a unique opportunity for the participants to familiarise themselves with the cutting-edge data infrastructure in the field of High Performance Computing (HPC). The school consisted of lectures and hands-on training on modern, GPU-accelerated supercomputers, GPU programming, GPU code optimization at scale, and understanding and applying machine learning methods in practice.

The school started with an inspiring Key Note Speech about Software Design and Performance Portability by Dr. Thomas Schultess, ETH Zurich/ CSCS. Various scientific case studies provided examples about using GPUs in different disciplines:
● Molecular dynamics on heterogeneous architectures: lessons learned from a decade of parallelization advances in GROMACS – Dr. Szilárd Páll, PDC, KTH Royal Institute of Technology
● Electronic-structure theory methods on GPUs – Dr. Jussi Enkovaara, CSC
● Weather and climate codes on GPUs – Dr. Daniel Klocke, MPI-M
● Program acceleration with GPU using CUDA – Dr. Talgat Manglayev, University of Helsinki, Department of Physics
● Computing intensive digital pathology: AI-driven cancer diagnostics achieves pathologist level performance – Asst. Prof. Pekka Ruusuvuori, University of Turku

In addition to learning new skills, the school enabled participants to widen their networks during frisbee golf, geocaching, and sauna evenings. Several of them even dared to swim in the cold lake!

Participants’ experiences

We interviewed some of the enthusiastic participants and asked about their experiences and motives. Anna Kormu (University of Helsinki, Finland) applied for the course to learn more about using OpenMP and GPUs in her work and doctoral thesis. She highlighted that this school was a valuable opportunity to learn how to use GPUs in high-performance computing. Ayush Chaturvedi (Virginia Tech, United States) is working on his thesis on optimization dedicated to different platforms and architectures. He found this school an excellent way to advance his studies.

For Manuel Stocchi (Research Institution – Italian Institute of Geophysics and Volcanology, Italy), the course was a good opportunity to travel to Finland and exploit the GPUs to see in practice if it was worth applying GPUs and OpenMP. He wanted to know if he could use GPUs to make better performance. It was an inspiring experience for him to exchange ideas related to the school topics and meet people worldwide. Ph.D. student Zhennan Kou (Aalto University, Finland) uses the lessons learned during the School for Molecular Dynamics. He wanted to learn how to improve the performance of GPUs. During the week, he also explored the nearby areas and visited towns close to Vuokatti.

For Ph.D. student Konstantinos Papadakis (University of Helsinki, Finland), this HPC event was a great chance to learn and set Vlasiator, a global hybrid Vlasov simulation code, to run on LUMI supercomputer. During the hackathon, he also managed to run and evaluate the current performance of Vlasiator’s GPU version. Also, Andreas Hadjigeorgiou (Cyprus Institute, Cyprus) found the school interesting for his Ph.D. One part of his research project is to create applications for different computer hardware. He also wanted to see if he could run his application on the new LUMI supercomputer, and eventually, he reached this goal as planned!

PRACE organises various training with different themes. To learn more about the upcoming PRACE Trainings, visit the PRACE Training Portal.

Author: Iiris Liinamaa, CSC