Parallelization and performance portability in hydrodynamics codes

With the eve of Exascale computing, performance and portability are atthe forefront of all scientific computing codes. Adding more CPU cores and moreenergy to a system is no longer a sustainable way to achieve improved perfor-mance, and extra effort must now be made to improve performance in all areas ofcode execution and code development. Using hydrodynamic simulation codes asa basis, this work explores numerous techniques to achieve performance in differ-ent ways. Adaptive mesh refinement (AMR) is a necessary technique to improvememory optimization in mesh-based simulations. However, it is invasive and con-ventionally difficult to integrate into existing applications, so we present a newbranch of AMR to create a smooth transition to these optimizations, which notonly improves performance, but also greatly reduces developer effort. We intro-duce the concept of this improvement as Phantom-Cell AMR, a new hybrid AMRscheme that aims to improve performance of AMR scientific application codes, aswell as provide a feasible way for existing applications to include AMR. Phantom-Cell AMR is tested with various architectures and parallel frameworks to exposethe optimizations it provides. Additionally, an application that uses this newAMR scheme to transition away from regular grid methods is investigated. Weshow that Phantom-Cell AMR is a way to adapt regular grid codes to includeAMR routines, and it can be done in an efficient, simple manner. Other workincluded in the document involves an investigation into an efficient data struc-ture that ensures optimal memory layout for cache performance, with a target ofmaking codes performant and portable across all architectures. All of the worktargets both performance and portability, not just on CPU hardware, but specifi-cally across GPU architectures. Parallel performance is key to all of the methodspresented, but the research makes a great effort to improve the portability of allapplications to prepare for current high performance computing (HPC) systemsand those on the horizon.