Domain Decomposition Methods in Science and Engineering XXVII

0583276 - MÚ 2024 RIV CH eng B - Monografie kniha jako celek
Dostál, Z. (ed.) - Kozubek, T. (ed.) - Klawonn, A. (ed.) - Langer, U. (ed.) - Pavarino, L. F. (ed.) - Šístek, Jakub (ed.) - Widlund, O. B. (ed.)
Domain Decomposition Methods in Science and Engineering XXVII.
1. - Cham: Springer, 2023. 528 s. Lecture Notes in Computational Science and Engineering, 149. ISBN 978-3-031-50768-7. ISSN 1439-7358
Institucionální podpora: RVO:67985840
Klíčová slova: Domain Decomposition Methods
Obor OECD: Applied mathematics
https://doi.org/10.1007/978-3-031-50769-4

This volume presents a selection of 62 peer-reviewed papers that were submitted to the proceedings of the 27th International Conference on Domain Decomposition Methods held in Prague, Czech Republic, from July 25 to 29, 2022.
With its first meeting in Paris in 1987, the International Conferences on Domain Decomposition Methods have been held in 16 countries in Asia, Europe, and North America, and now for the first time in the Czech Republic. The conference is held at roughly 18-month intervals. A complete list of the 27 meetings appears below. Domain decomposition is often seen as a formof the divide-and-conquer approach for mathematical problems posed over a physical domain, reducing a large problem into a collection of smaller problems, each of which is much easier to solve computationally
than the undecomposed problem, and most or all of which can be solved independently and concurrently, and then solved iteratively in a consistent way. A lot of the theoretical interest in domain decomposition algorithms lies in ensuring that the number of iterations required to converge is very small. Domain decomposition algorithms can be tailored to the properties of the physical system, as reflected in the mathematical operators, to the number of processors available, and even to specific architectural parameters, such as cache size and the ratio of memory bandwidth to floating point processing rate. Consequently, domain decomposition methods prove to be an ideal paradigm for large-scale simulation on advanced parallel computers and supercomputers. While the technical content of the conference revolves mainly around mathematics, its underlying motivation lies in enabling efficient utilization of distributed memory computers for complex scientific and engineering applications. Although research on domain decomposition methods is presented at various events, the International Conference on Domain Decomposition Methods stands as the singular recurring international forum dedicated to fostering interdisciplinary interactions between theoreticians and practitioners. These interactions span the development, analysis, software implementation, and applications of domain decomposition methods. As we are entering the era of exascale computing, with the most powerful supercomputers now capable of sustaining 1018 floating-point operations per second, the need for efficient and mathematically sound methods for solving large-scale systems becomes increasingly vital. Furthermore, these methods must align well with the modern high-performance computing (HPC) architectures. The massive parallelism inherent in exascale computing necessitates the development of new solution methods that effectively leverage the abundance of computing cores and hierarchical memory access patterns. Ongoing advancements, such as parallelization in time, asynchronous iterative methods and nonlinear domain decomposition methods show that this massive parallelism not only calls for novel solution and discretization approaches but also facilitates their further development.
Trvalý link: https://hdl.handle.net/11104/0351274