We present an upgrade to the particle-in-cell ion beam simulation code OPAL that enables us to run highly realistic simulations of the spiral inflector system of a compact cyclotron. This upgrade includes a new geometry class and field solver that can handle the complicated boundary conditions posed by the electrode system in the central region of the cyclotron both in terms of particle termination, and calculation of self-fields. Results are benchmarked against the analytical solution of a coasting beam. As a practical example, the spiral inflector and the first revolution in a 1 MeV/amu test cyclotron, located at Best Cyclotron Systems, Inc., are modeled and compared to the simulation results. We find that OPAL can now handle arbitrary boundary geometries with relative ease. Comparison of simulated injection efficiencies, and beam shape compare well with measured efficiencies and a preliminary measurement of the beam distribution after injection. * winklehn@mit.edu; Massachusetts Institute of Technology,

Significant problems facing all experimental and computational sciences arise from growing data size and complexity. Common to all these problems is the need to perform efficient data I/O on diverse computer architectures. In our scientific application, the largest parallel particle simulations generate vast quantities of six-dimensional data. Such a simulation run produces data having an aggregate data size up to several TB per run. Motived by the need to address data I/O and access challenges, we have implemented H5Part, an open source data I/O API that simplifies the use of the Hierarchical Data Format v5 library (HDF5). HDF5 is an industry standard for high performance, crossplatform data storage and retrieval that runs on all contemporary architectures from large parallel supercomputers to laptops. H5Part, which is oriented to the needs of the particle physics and cosmology communities, provides support for parallel storage and retrieval of particles, structured and in the future unstructured meshes. In this paper, we describe recent work focusing on I/O support for particles and structured meshes, and provide data showing performance on modern supercomputer architectures like the IBM POWER 5. MOTIVATION

No abstract

Abstract-Particle-based simulations running on large highperformance computing systems over many time steps can generate an enormous amount of particle-and field-based data for post-processing and analysis. Achieving high-performance I/O for this data, effectively managing it on disk, and interfacing it with analysis and visualization tools can be challenging, especially for domain scientists who do not have I/O and data management expertise. We present the H5hut library, an implementation of several data models for particle-based simulations that encapsulates the complexity of HDF5 and is simple to use, yet does not compromise performance.

No abstract