A data-driven approach for discontinuous integration
Please login to view abstract download link
We present a data-driven approach for the integration of discontinuous functions and use the finite cell method as model problem. The accuracy and performance of the proposed method is tested on state-of-the-art accelerator hardware such as Nvidia A100 GPU cards and compared to standard integration methods such as adaptive quadrature. The data-driven approach provides constant computational complexity, independent of the cut configuration, and offers an efficient alternative that can take advantage of the computational power of accelerators.
