Research

Numerical Methods

Our numerical methods research focuses on developing efficient and accurate algorithms for solving complex mathematical problems. We work on both theoretical foundations and practical implementations, with particular emphasis on high-performance implementations of numerical methods for state-of-the-art supercomputers, including integrations with popular application libraries.

Current research topics include:

• Sparse-primitives: SpMV, SpGEMM
• Iterative and direct linear system solvers
• Sophisticated preconditioners for linear systems

Scientific Computing

Our scientific computing research enables simulation of complex physical systems through high-performance computing. We develop software frameworks and algorithms that leverage modern hardware architectures for maximum efficiency.

Current research topics include:

• Parallel algorithms for various applications: CFD, Electrophysiology, combustion, Plasma physics
• GPU-accelerated scientific computing

Research Software

We are committed to building high-performance open-source software.

Some research software that we actively develop are:

Ginkgo

Ginkgo is a high-performance linear algebra library for manycore systems, with a focus on solution of sparse linear systems.

NeoN

Prototype of a modern CFD core

Funded Projects

Our research is supported by various national and international funding agencies.

Active Projects

WarmWorld

Capitalizing on recent advances in HPC, WarmWorld centrally contributes to this new paradigm of high-resolution models by furthering the development of ICON, a globally coupled climate model. To this end, WarmWorld’s efforts are not concentrated on constructing a model from scratch. The project rather imparts a significant contribution to ongoing developments of the already convection-resolving ICON model. Overall, WarmWorld aims to increase ICON’s run time efficiency, construct a more fine-meshed model grid, and seeks to create a more seamless experience for end-users. The leap toward smaller scales allows for the representation of finer meso-scale processes, such as cloud formation and eddies, and holds not only with the promise to improve local impact assessments but fosters potential to improve understanding of large-scale processes at the same time.

MICROCARD-2

MICROCARD builds software that can simulate cardiac electrophysiology using whole-heart models with sub-cellular resolution, on exascale supercomputers. We are part of work-package 2, which aims to deliver efficient large-scale distributed preconditioners and solvers.