Developing numerical methods for high-performance computing on supercomputers is the focus of the research conducted by Anzt, Professor of Computational Mathematics at the TUM School of Computation, Information and Technology (CIT). Because making these algorithms available to the research community is important to him, Anzt contributed to the development of GINKGO. He explains: “GINKGO is a software package that delivers diverse numerical methods for solving equation systems. It basically works like a toolbox the community can dig into to solve equations.”
Today, GINKGO is used by a wide array of organizations for simulations in a number of application settings. “In most cases, we do not find it relevant what exactly is to be calculated. For every request to solve an equation system of a specific size with specific characteristics we can point to the respective tool that can be employed,” says Anzt. For example: “The Princeton Plasma Physics Laboratory uses GINKGO to simulate fusion reactors, and the Lawrence Livermore National Laboratory resorts to our library for its rocket propulsion simulations.”
Understanding the Human Heart Better
Another example of GINKGO applications is MICRO-CARD, a project created by the European High-Performance Computing Joint Undertaking (EuroHPC JU). “The undertaking is about simulating the human heart, more precisely, the electrochemical processes in it,” the computer scientist explains. “We know individual heart cells communicate with each other using electrochemical processes so that they beat in a synchronized manner. What we don’t understand yet is the exact processes, how it works, and why arrythmias or heart attacks occur.” The reason simulations of the human heart at the cell level have failed so far is that the human heart contains billions of cells, and the computing power has been insufficient for such an elaborate model. Therefore, understanding the processes at the cell level requires new numerical methods and a scalable software architecture for super-computers. Anzt says: “Our goal is to detect arrythmias early on and develop effective treatments by identifying and potentially removing problematic cells.”
More Accurate Weather Forecasts and Climate Predictions
Anzt and his team also offer their expertise for an entirely different type of application: They are working to accelerate the ICON climate code by means of efficient software and numerical methods and to optimize its implementation in diverse systems. Developed by Germany‘s National Meteorological Service, Max Planck Institute, the Karlsruhe Institute of Technology (KIT), and other institutions, ICON is used, among others, by the German Climate Computing Center and for daily weather forecasts presented in the Tagesschau news program.
ICON uses a grid that maps the Earth to facilitate weather forecasts for areas that are as narrowly defined as possible. The finer the grid, the more accurate the simulation and the more complex the computing. Creating such detailed grids currently is very time-consuming. For this reason, Anzt and his team focus on speeding up the simulations.
Another goal is to make ensemble simulations easier in order to reduce uncertainties in weather forecasts and climate predictions. The researchers conduct a series of simulations, each one with some slightly altered values, to observe the impact on the probability of rain, for example. “For the ensembles we need to either boost computing power or make the individual simulations more efficient. That is what we are working on,” Anzt adds.
Relevant to a Number of Stakeholders
A variety of stakeholders could benefit from accurate weather forecasts and climate predictions. According to Anzt, financial interests come first. He gives an example: “If I know whether the sun will shine three days from now, I can buy or sell electricity on the stock market accordingly.” He continues by saying that naturally climate models also have political relevance because they provide information on how specific regions will develop over the next several decades, if they are more likely to experience extreme weather events, and if they will remain habitable in the future – key factors for long-term measures and planning.
History has shown the relevance of precise weather forecasts even to the military: After a brief period of good weather was announced for June 6, 1944, the allies in World War II decided to land their troops in Normandy on this exact day. The weather held, the operation was successful, and the day went down in history as D-Day.