Numerical simulations on High Performance Computing (HPC) systems are essential tools in astrophysics and cosmology today. They are theoretical laboratories in which we can investigate, interpret and understand the physical processes behind the observed sky. Astrophysical systems often involve highly non-linear and interconnected processes over a huge range of dynamical scales. Take galaxy formation as an example, galaxy evolution is strongly influences by large-scale processes like the accretion of intergalactic gas (at Mega-parsec scale), and at the same time star formation, black hole formation deep within the galaxy (within parsec scales). Feedback from supermassive black holes can impact significantly the entire galaxy and intergalactic space. To be able to resolve and understand the interconnectivity of processes in a numerical simulation is crucial for providing robust models and interpretations for observations. With the development of exascale HPC systems, it is important that astrophysical codes are able to exploit these systems efficiently. However, exascale systems are expected to have a heterogeneous and unprecedented architectural complexity, which has a significant impact on simulation codes. The main objective of the SPACE CoE is to extensively re-engineer eight widely-used astrophysics codes and to enable them to harness exacale computing capabilities. This will be accomplished by co-design activities and interdisciplinary collaborations between scientists, code developers, HPC experts and hardware manufacturers. In addition, SPACE will address the high-performance data analysis of the data torrent produced by exascale astrophysical simulations, also with machine-learning and visualisation tools. The Norwegian partner is responsible for co-designing of the ChaNGa code, which is widely used in cosmology and galaxy formation.
In Astrophysics and Cosmology (A&C) today, High Performance Computing (HPC)-based numerical simulations are outstanding instruments for scientific discovery. They represent essential tools and theoretical laboratories able to investigate, interpret and understand the physical processes behind the observed sky. For these laboratories, the efficient and effective exploitation of exascale computing capabilities is essential. Exascale systems, however, are expected to have a heterogeneous unprecedented architectural complexity, with a significant impact on simulation codes. Consequently, the proposed SPACE CoE aims to extensively re-engineer the target codes to engage with new computational solutions and adopt innovative programming paradigms, software solutions, and libraries. SPACE aims to foster the reuse and sharing of algorithms and software components in the A&C application domain. The proposed SPACE CoE will address this action through co-design activities that bring together scientists, code developers, HPC experts, HW manufacturers and SW developers, advancing lighthouse exascale A&C applications, and promoting the use of upcoming exascale and post-exascale computing capabilities. In addition, SPACE will address the high-performance data analysis of the data torrent produced by exascale A&C simulation applications, also with machine-learning and visualisation tools. The deployment of applications running on different platforms will be facilitated by federating capabilities focusing on code repositories and data sharing, and integrating European astrophysical communities around exascale computing by adopting software and data standards and interoperability protocols.
The Norwegian partner will participate in exascale enabling, performance analysis, optimising and co-designing of one A&C code (ChanGa) together with other science and industrial partners. We will also address extreme data processing and analysis, with machine-learning and visualisation tools.