Advanced parallel-in-time algorithms for computer simulations in physical sciences, social sciences and engineering
Date: 20 - 24 May 2019
Convenors: Daniel Ruprecht (Leeds, GBR), Robert Speck (Jülich, GER), Sebastian Schöps (Darmstadt, GER)
The workshop will bring together developers and users of advanced parallel mathematical algorithms for computer simulations. Computer simulations are heavily used in all areas of the physical sciences and are becoming increasingly important in the social sciences. Because most disciplines use models based on differential equations, the used algorithms are often similar despite the differences in the modelled application. Finite difference methods, for example, are used in fluid dynamics, to model CNC machines in industrial manufacturing or to simulate formation of black holes. The complexity of many real-world problems means that often high-performance computing is required to run simulations with acceptable solution times. However, the trend towards extreme parallelism makes developing the necessary numerical algorithms challenging. More than ever, the development of efficient algorithms requires close collaboration across disciplines. Applied mathematicians, computer scientists and domain scientists who are experts in the targeted application all need to be involved to deliver algorithms that are robust, accurate, efficient on massively parallel computers and effective in capturing all the relevant features of the studied problem. The workshop will bring these groups together to foster exchange, stimulate innovation and to explore new areas, particularly in the computational social sciences, where modern parallel computing algorithms could find an application.