Algebraic multigrid preconditioning for computational fluid dynamics on high performance computers

In Computational Fluid Dynamics (CFD), sparse iterative solvers are widely em- ployed to efficiently compute solutions to large-scale problems. The choice of an appropriate preconditioning technique plays a pivotal role in accelerating con- vergence and reducing computational time. Among these techniques, Algebraic Multigrid (AMG) has emerged as a state-of-the-art approach for preconditioning, particularly effective for large and complex systems. This thesis presents a perfor- mance analysis of different software implementations of AMG, focusing on their efficiency and scalability when applied to real-world problems on different high- performance computing architectures. The results provide valuable insights into optimizing solver performance for CFD applications.