Slides: What Happens in MPI Shouldn't Stay in MPI: Making Communication Visible for Learning and Performance Analysis

Slides of the invited talk at EuroMPI 2026, TU Wien, Vienna, Austria, October 8, 2026.Speaker: Anna-Lena Roth, Fulda University of Applied Sciences, Germany. Learning MPI means more than learning its syntax. Efficient parallel programs require understanding why communication is necessary but expensive, why an exchange within a node differs from one across the network, and which performance problems arise from communication itself. Such understanding is part of performance analysis, which typically appears late in curricula. Collective operations illustrate this: learners see a single call, while the communication actually carried out by the MPI library remains invisible. The talk presents approaches to teaching MPI in which communication itself becomes the central representation. Based on three years of classroom experience with EduMPI Suite, it covers: the situation of HPC education in Germany, where few universities of applied sciences have a cluster and performance analysis is rarely taught in practice; the move from performance analysis as a last-week add-on to first-day practice; the near-real-time visualization of MPI communication, mapped onto processes, nodes and cores; the decomposition of collective operations into their underlying point-to-point communication, which makes the selected algorithm observable; the limits of PMPI and MPI_T for obtaining this information, and why EduMPI is currently a fork of Open MPI; results of a comparative usability study, in which 96 % of analysis tasks were solved with EduMPI Suite compared with 60 % using TAU or CUBE; what academia and the MPI standard could contribute to make MPI communication accessible to learning and performance analysis tools; an outlook on observing asynchronous many-task (AMT) runtimes that use MPI as their backend. Software EduMPI (Open MPI fork for near-real-time MPI performance measurement, including EduStore setup): https://github.com/AnnaLena77/EduMPI EduMPI GUI (cross-platform interface for running MPI programs on clusters and visualizing communication): https://github.com/AnnaLena77/EduMPI_GUI Related publications Roth, A.-L.; Süß, T. (2023): Performance Analysis Tools for MPI Applications and their Use in Programming Education. ICPE '23 Companion, ACM, pp. 361–368. https://doi.org/10.1145/3578245.3584358 Roth, A.-L.; James, D.; Kuhn, M.; Konert, J. (2024): Enhancing Parallel Programming Education with High-Performance Clusters Utilizing Performance Analysis. DELFI 2024, Gesellschaft für Informatik e.V. https://doi.org/10.18420/delfi2024_42 Roth, A.-L.; James, D.; Kuhn, M. (2025): EduMPI – Simplifying the Use of High-Performance Clusters and Focusing Performance Analysis in Parallel Programming Education. PARS-Mitteilungen, Vol. 37. https://dl.gi.de/handle/20.500.12116/46131 Roth, A.-L.; James, D.; Kuhn, M.; Frisch, D. (2025): Making MPI Collective Operations Visible: Understanding Their Utility and Algorithmic Insights. Euro-Par 2025, LNCS 15900, Springer. https://doi.org/10.1007/978-3-031-99854-6_5 Roth, A.-L.; James, D.; Posner, J.; Kuhn, M. (2026): Performance Analysis in Parallel Programming Education: A Comparative Usability Study. PARS-Mitteilungen, Vol. 38. https://dl.gi.de/handle/20.500.12116/48841 Roth, A.-L.; Posner, J.; Kuhn, M. (2026): Enabling Interactive Visualization for Near-Real-Time Performance Analysis of MPI Applications. ICPE '26 Companion, ACM, pp. 120–129. https://doi.org/10.1145/3777911.3800642 Roth, A.-L.; Posner, J. (2026): An Empirical Analysis of High-Performance Computing Education in Germany. ICCS 2026 Workshops (WTCS), LNCS 16789, Springer. https://doi.org/10.1007/978-3-032-29918-5_32 Roth, A.-L.; Posner, J. (2026): From a Last-Week Add-On to First-Day Practice: Three Years of Integrating MPI Performance Analysis into HPC Education with EduMPI Suite. EduHPC, SC '26 Workshops (to appear). Preprint: https://doi.org/10.48550/arXiv.2610.06352 Roth, A.-L.; Posner, J. (2026): Near-Real-Time Performance Visualization of MPI Applications: Integrating Performance Analysis into HPC Education (Poster, to appear). Doctoral Showcase, SC26: The International Conference for High Performance Computing, Networking, Storage, and Analysis, Chicago, IL, USA. https://doi.org/10.5281/zenodo.23059851 Datasets End-of-Course Student Surveys on MPI Performance Analysis with TAU in a Master's Parallel Programming Course (2026). https://doi.org/10.5281/zenodo.22662192 Usability Evaluation EduMPI 2025 (2025). https://doi.org/10.5281/zenodo.16798981