IS, Beijing, People's Republic of China
IMP, Lanzhou, People's Republic of China
ANL, Argonne, USA
Electromagnetic simulations are fundamental for accelerator modeling. In this paper two high-order numerical methods will be studied. These include continuous Galerkin (CG) method with vector bases, and discontinuous Galerkin (DG) method with nodal bases. Both methods apply domain decomposition method for the parallelization. Due to the difference in the numerical methods, these methods have different performance in speed and accuracy. DG method on unstructured grid has the advantages of easy parallelization, good scalability, and strong capability to handle complex geometries. Benchmarks of these methods will be shown on simple geometries in detail first. Then they will be applied for simulation in accelerator devices, and the results will be compared and discussed.
ANL, Argonne, USA
RPI, Troy, New York, USA
We examined hybrid parallel infrastructures in order to ensure performance and scalability for beam propagation modeling as we move toward extreme-scale systems. Using an MPI programming interface for parallel algorithms, we expanded the capability of our existing electromagnetic solver to a hybrid (MPI/shared-memory) model that can potentially use the computer resources on future-generation computing architecture more efficiently. As a preliminary step, we discuss a hybrid MPI/OpenMP model and demonstrate performance and analysis on the leadership-class computing systems such as the IBM BG/P, BG/Q, and Cray XK6. Our hybrid MPI/OpenMP model achieves speedup when the computation amounts are large enough to compensate the OMP threading overhead.
