On June 25th 2021, POP partners (RWTH Aachen University, Barcelona Supercomputing Center), together with the University of Arizona and the Federal University of Rio Grande do Sul, organized a virtual workshop PERMAVOST.
This workshop was created to foster collaborations with stakeholders in performance research, analysis and engineering ecosystems. Modern software engineering is getting increasingly complicated, especially in the HPC field, where the complication not only comes from working with cutting-edge technology and novel problems, but also from the scale of the problem and from the team itself. We are increasingly working with stakeholders of various backgrounds and varying degrees of expertise in HPC performance analysis. Bridging these diverse stakeholders and creating a robust feedback loop can be a way to prepare for future challenges in performance modelling, analysis, engineering and visualization.
The workshop gathered more than twenty attendees, from as far away as the US west coast and Japan. The workshop's call for papers attracted submissions from Barcelona Supercomputing Center, Chiba Institute of Technology and the Technical University of Munich. Besides the paper presentations, there was a keynote speech from Dr. Andreas Herten from Juelich Supercomputing Center (JSC). Dr. Herten shared his experience managing JUWELS Boosters Early Access Program. JUWELS Booster is a recently installed Tier-0/1 system at JSC and is currently the seventh fastest supercomputer in the world and the fastest in Europe. In this talk, participants gained insights into the state of the art in the performance analysis and modelling of GPUs from JUWELS Booster Early Access Program.
The last event of the workshop was a discussion panel, with panellists from industry (Dr. Max Katz from NVIDIA), academia (Dr. Lucas M. Schnorr from the Federal University of Rio Grande do Sul), and research institutions (Dr. Guido Juckeland from Helmholtz-Zentrum Dresden-Rossendorf and Dr. Olga Pearce from Lawrence Livermore National Laboratory). The discussion covered topics related to the future of performance analysis techniques and tools, with specific regard to how GPUs and Cloud Computing applications are changing a lot of approaches in performance analysis and engineering.
The workshop proceedings are available in the ACM digital library:
https://dl.acm.org/doi/proceedings/10.1145/3452412