Session

DescriptionSolving Linear Algebra problems is at the core of many scientific and engineering applications. When the problem size is growing, it is critical to maintaining a reasonable run-time while using the computational resources at full capacity. The need for scalable algorithms is essential in preparing scientific applications for the new exascale ecosystem. Having flexibility and efficiency for the related software framework on supercomputers with hybrid nodes lies at this effort's core. This mini-symposium will address the development of libraries of solvers and preconditioners for the applications related to the current European transition towards renewable energies. In particular, in the context of the "Energy Oriented Center of Excellence: toward exascale for energy" European project on High-Performance Computing and Sustainable Energies.
The Talks will cover topics such as the acceleration and scalability techniques for direct solvers as the Block-Low Rank (BLR) feature or the use of GPUs. Recent developments in Multigrid Solvers, both Algebraic and Geometric type, will be discussed, e.g., the usage of graph matching aggregation schemes and direct distributed coarse solvers with BLR features.
The talks will explore several applications ranging from the simulation of hydrodynamic systems and large wind and material science models to illustrate these approaches' performance.