Minisymposium Presentation

In recent years, high-performance computing (HPC) systems have experienced exponential growth in performance that has outpaced the improvements in energy efficiency. While this progress has significantly advanced many scientific fields, like drug discovery, fusion reactor modeling, and climate research, it has also led to substantial energy consumption, contributing to significant carbon dioxide emissions. Addressing this environmental impact is essential to align HPC development with Sustainable Development Goals (SDGs), particularly those focusing on responsible consumption . This study explores energy consumption in particle simulations, with examples from applications in Molecular Dynamics (MD). The analysis highlights challenges in optimizing energy efficiency across hardware and for many different simulation scenarios. A solution to tackle these challenges is presented in the form of AutoPas, a particle simulation library that dynamically selects the optimum algorithm to minimize runtime or energy use. Some strategies to improve energy efficiency include reducing memory-intensive algorithms, leveraging low-level parallelism, etc. are discussed, and examples from MD simulations are presented. This talk advocates for the importance of sustainable approaches in HPC development, that is to mitigate environmental impacts of computing while still ensuring the much needed progress in science and in turn, to improving human lives.