Accurately Measuring Energy Consumption of Large Cosmological Simulations
DescriptionSPH-EXA is a highly scalable and extendable simulation framework for astrophysical and cosmological simulations, codesigned by computational scientists and computer scientists to achieve scientific advances in astrophysics, cosmology, and high-performance computing. This talk discusses extending the SPH-EXA framework by integrating power measurement tools to accurately measure the energy usage of each simulation function during the entire simulation. The collected data enables the energy efficiency analysis of large simulations allowing corrective measures to be implemented. We show the power, energy and carbon footprint data collected on multiple large simulations for hydro only, hydro+gravity and hydro+gravity coupled with cooling test cases analyzing the energy profile of each test case to find the trade-off points between performance and power consumption.
TimeWednesday, June 2814:00 - 14:30 CEST
Computer Science, Machine Learning, and Applied Mathematics