Applicant
Prof. Dr.-Ing. Wolfgang A. Wall
Institute for Computational Mechanics
Technical University of Munich
Project Overview
We propose an extensive performance optimization of the core infrastructure of the 4C code (formerly known as BACI). Over its 20-year history, this unique multiphysics research code has been used in hundreds of scientific contributions to simulate complex real-world applications. How- ever, technological advances in hardware nowadays mandate different software implementations and code architecture to fully exploit computational resources. Significant contributions to high- performance computing have been made by the institute within the deal.II library, enabling efficient large-scale simulations. The current proposal aims to take the lessons learned in these projects and enable a leap in the performance of the 4C code base. As a first step, the finite element (FE) discretization in 4C needs to be overhauled based on our great experience with special features in the deal.II library. Once this core feature is realized, the efficiency of evaluation of domain integrals – omnipresent in FE – will be improved, aiming to demonstrate at least a 10x performance boost in practical examples. Furthermore, using the established mesh representation of deal.II provides the opportunity to explore adaptive mesh refinement. All work carried out within this project will be made available as open-source code.