The course offers an in-depth exploration of advanced numerical techniques for efficiently solving large-scale sparse linear systems, with a particular focus on leveraging the capabilities of the Parallel Sparse Computation Toolkit (PSCToolkit). PSCToolkit is a comprehensive suite of libraries designed for high-performance computing environments, enabling the solution of symmetric and positive-definite linear systems using thousands of GPUs and tens-of-thousands cores on current supercomputers. The toolkit encompasses components such as PSBLAS for parallel sparse basic linear algebra operations, and AMG4PSBLAS for algebraic multigrid preconditioners . Through this course, participants will gain practical experience in implementing iterative solvers, implementing and utilizing algebraic multilevel preconditioners, and optimizing performance on distributed-memory architectures. The curriculum is designed to equip students with the skills necessary to both tackle computational challenges in fields like computational fluid dynamics, structural analysis, energy simulations, and to show directions and possible activities for research in innovative iterative and preconditioning methods.
Lectures will be held by Salvatore Filippone from the University of Rome ``Tor Vergata''.