OPALX: A performance-portable framework for particle accelerator simulations
Abstract
Accurate charged-particle beam simulations require self-consistent collective fields and realistic accelerator models, while their computational cost motivates efficient use of heterogeneous computing systems. We present OPALX, a performance-portable accelerator simulation framework that builds on the beam-dynamics capabilities of OPAL. Its modular architecture uses the Independent Parallel Particle Layer (IPPL) and Kokkos to support distributed-memory execution on multicore CPUs and GPUs from a common codebase. The framework combines relativistic particle tracking with prescribed accelerator fields, particle-in-cell space charge calculations, and spin tracking. We describe the physical and numerical models and their integration into the simulation workflow. The implementation is assessed using disorder-induced heating and frozen-spin test cases, together with comparisons of spin evolution with G4Beamline and beam transport with TraceWin. A three-dimensional space charge benchmark evaluates throughput on CPU and GPU systems, reaching approximately 109 macroparticles per second on a single NVIDIA A100 GPU for this benchmark. These results provide initial verification of the implemented models and demonstrate CPU and GPU execution of self-consistent space charge calculations. OPALX provides a common foundation for extending accelerator models and numerical methods across heterogeneous computing platforms.
Related
- A HIP-Compatible Accelerator Backend for Fourier-Bessel Particle-in-Cell Simulations on CPU/DCU Heterogeneous Clusters
- Performance-portable GPU acceleration of the hybrid particle-in-cell code dHybridR
- KEMPIC-3D: A transparent and extensible electromagnetic Particle-in-Cell framework for kinetic plasma simulations
- NEOPIC: A Neural Operator Framework for Particle-based Kinetic Plasma Simulations
- Brilliance: toward autonomous computational research in particle accelerator physics
BibTeX
@misc{zampetakis2026opalx,
title = {{OPALX: A performance-portable framework for particle accelerator simulations}},
author = {Michail Zampetakis and Ryan Sean Ammann and Yuanjie Bi and John Biddiscombe and Pedro Calvo and Carl Jolly and Severin Andrey Klapproth and Alexander Paul Liemen and Sriramkrishnan Muralikrishnan and Chris Rogers and Mohsen Sadr and Jonathan Thompson and Andreas Adelmann},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.00124},
eprint = {2611.00124},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.00124},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - OPALX: A performance-portable framework for particle accelerator simulations
AU - Michail Zampetakis
AU - Ryan Sean Ammann
AU - Yuanjie Bi
AU - John Biddiscombe
AU - Pedro Calvo
AU - Carl Jolly
AU - Severin Andrey Klapproth
AU - Alexander Paul Liemen
AU - Sriramkrishnan Muralikrishnan
AU - Chris Rogers
AU - Mohsen Sadr
AU - Jonathan Thompson
AU - Andreas Adelmann
PY - 2026
DA - 2026/11/01
PB - arXiv
DO - 10.48550/arxiv.2611.00124
UR - https://arxiv.org/abs/2611.00124
AB - Accurate charged-particle beam simulations require self-consistent collective fields and realistic accelerator models, while their computational cost motivates efficient use of heterogeneous computing systems. We present OPALX, a performance-portable accelerator simulation framework that builds on the beam-dynamics capabilities of OPAL. Its modular architecture uses the Independent Parallel Particle Layer (IPPL) and Kokkos to support distributed-memory execution on multicore CPUs and GPUs from a common codebase. The framework combines relativistic particle tracking with prescribed accelerator fields, particle-in-cell space charge calculations, and spin tracking. We describe the physical and numerical models and their integration into the simulation workflow. The implementation is assessed using disorder-induced heating and frozen-spin test cases, together with comparisons of spin evolution with G4Beamline and beam transport with TraceWin. A three-dimensional space charge benchmark evaluates throughput on CPU and GPU systems, reaching approximately 10^9 macroparticles per second on a single NVIDIA A100 GPU for this benchmark. These results provide initial verification of the implemented models and demonstrate CPU and GPU execution of self-consistent space charge calculations. OPALX provides a common foundation for extending accelerator models and numerical methods across heterogeneous computing platforms.
ER -