GPU Computing and Krylov Solvers

Explore GPU computing and Krylov solvers in this comprehensive 1-hour 29-minute lecture from the Advanced Cyberinfrastructure Training at RPI. Delve into the intricacies of leveraging Graphics Processing Units (GPUs) for high-performance computing applications, with a specific focus on implementing Krylov subspace methods. Learn how to harness the parallel processing power of GPUs to accelerate linear algebra operations and iterative solvers commonly used in scientific computing and engineering simulations. Gain insights into optimizing algorithms for GPU architectures, understanding memory hierarchies, and efficiently utilizing CUDA or other parallel programming frameworks. Discover techniques for improving the performance of Krylov solvers such as Conjugate Gradient, GMRES, and BiCGSTAB on GPU platforms, and explore case studies demonstrating the significant speedups achievable in large-scale numerical simulations.

GPU Computing and Krylov Solvers