Modern applications - ranging from large-scale graph analytics to machine learning and combinatorial optimization - demand algorithms that are not only asymptotically efficient but also optimized for massive parallelism and/or complex memory hierarchies. However, the theoretical algorithm design community and the systems-level Software Performance Engineering (SPE) community frequently operate in silos, creating a widening gap between theoretical potential and practical implementation. This seminar aims to convene researchers in both communities and bridge this gap by establishing a rigorous, unified discipline that spans the entire lifecycle of algorithm development, from principled design to empirical evaluation.

This Dagstuhl Seminar addresses critical interdisciplinary challenges: translating theoretical insights into practical high-performance software, establishing shared standards for reproducible benchmarking, and developing performance-portable algorithms across diverse platforms (e.g., CPUs, GPUs, and HPC clusters). Beyond technical methodology, the discussions will focus on building long-term community infrastructure. Furthermore, the seminar will explore bidirectional technology transfer between academia and industry, as well as the development of educational resources that integrate algorithmic principles with systems engineering, which ultimately equip a new generation of researchers to tackle the next frontier of high-performance computing.

To maximize collaborative impact, the seminar will move beyond traditional formal presentations to an interactive environment. Through facilitated thematic deep-dives, organized working groups, and open problem sessions, participants will identify shared techniques and recurring challenges across domains. The goal is to produce concrete outcomes - such as shared research roadmaps, benchmark specifications, and educational curricula - that will solidify the connection between algorithm engineering and SPE, enabling theoretical breakthroughs to become the high-performance building blocks of tomorrow.

Creative Commons BY 4.0

José Nelson Amaral, David A. Bader, Petra Mutzel, and Yihan Sun