GPU Acceleration of Optical Photon Propagation Achieves 161x Speedup for Electron Ion Collider Simulations
Summarize this article with:
Simulating the detection of particles often relies heavily on accurately modelling how light, in the form of photons, travels through complex detectors, a process that demands significant computational resources. Gabor Galgoczi, Kolja Kauder, Maxim Potekhin, and their colleagues at Brookhaven National Laboratory have developed EIC-Opticks, a new approach to accelerate these simulations, particularly for experiments expected at the future Electron-Ion Collider. This innovative software leverages the power of graphics processing units (GPUs) and a technique called event aggregation to dramatically speed up the tracking of photons, achieving speedups of up to an order of magnitude compared to conventional methods. By streamlining this crucial step in detector simulation, EIC-Opticks promises to enable more detailed and efficient modelling of particle physics experiments, ultimately enhancing our ability to analyse data and uncover new insights into the fundamental laws of nature. Leveraging the power of graphics processing units (GPUs) and NVIDIA OptiX, EIC-Opticks speeds up calculations traditionally performed using CPU-based methods like GEANT4. A key innovation is event aggregation, which processes thousands of individual events simultaneously, dramatically reducing simulation time and achieving a 100-fold increase in throughput by processing 10,000 events in a single batch.
The team validated EIC-Opticks by simulating the ePIC pfRICH detector, demonstrating its ability to accurately reproduce results obtained with GEANT4. The framework also includes tools to manage complex detector geometries and minimize photon leakage, a common issue in ray tracing, and provides scripts to optimize settings for specific detector configurations. By accelerating these simulations, scientists can design and optimize detectors more quickly, conduct larger-scale studies of detector performance, and generate larger training datasets for machine learning applications. EIC-Opticks represents a powerful tool that will significantly advance the development and understanding of detectors for the EIC and other physics experiments, with potential applications in medical imaging and other fields. GPU Acceleration of Photon Transport Simulation EIC-Opticks represents a significant advancement in simulating the behaviour of photons within detectors used in high-energy physics experiments. Researchers have developed this new framework, building upon existing GPU-accelerated techniques, to dramatically speed up the computationally intensive task of modelling optical photon transport. By employing event aggregation, EIC-Opticks batches thousands of individual low-yield events for simultaneous processing on graphics processing units, achieving speedups of up to an order of magnitude compared to conventional multi-threaded simulations and up to 161times faster than single-threaded execution. This acceleration is particularly beneficial for experiments like the ePIC pfRICH detector, where simulating photon behaviour is crucial for accurate data analysis.
The team also addressed practical challenges in detector simulation, implementing safeguards against common errors arising from complex detector geometries. EIC-Opticks analyses detector designs and warns users about potentially problematic configurations, preventing crashes and ensuring reliable results. Furthermore, the researchers have prioritised usability and reproducibility by providing easy installation options through both Spack packages and Docker containers, alongside tools for optimising key simulation parameters. Validation using the ePIC pfRICH detector demonstrates excellent agreement between GPU and CPU simulations, confirming the accuracy of the new approach. For low-to-moderate photon yield experiments, event aggregation reduces the per-photon simulation time from 60 microseconds to approximately 20 nanoseconds with batching, representing a 3000-fold improvement. Experiments reveal that EIC-Opticks delivers a factor of 10 speedup compared to multi-threaded Geant4 simulations. GPU Acceleration of Photon Transport Simulation Researchers have developed EIC-Opticks, a novel GPU-accelerated framework that significantly enhances the simulation of optical-photon transport, a computationally intensive task crucial for detector studies in high-energy and nuclear physics. Building upon the existing Opticks software, EIC-Opticks couples to Geant4, offloading optical physics to NVIDIA’s OptiX ray-tracing runtime while retaining all other physics calculations on the CPU. The core innovation lies in an event-aggregation mode, which batches thousands of individual low-yield events, typically producing between 100 and 10,000 photons per event, into a single GPU simulation. This approach eliminates the substantial per-event launch and data transfer overheads that previously limited the efficiency of GPU acceleration for experiments with modest photon yields.
The team validated EIC-Opticks using the pfRICH detector as a representative example of an Electron Ion Collider subsystem, conducting simulations with 50,000 electrons at a momentum of 5 MeV/c. Comparisons between Geant4 and EIC-Opticks demonstrate excellent agreement in observable quantities, confirming the fidelity of the new method. Performance benchmarks reveal substantial speedups; the batched EIC-Opticks framework achieves speedups exceeding two orders of magnitude compared to single-threaded Geant4, and outperforms multi-threaded Geant4 by a factor of ten for typical pfRICH workloads. To ensure usability and reproducibility, scientists provide EIC-Opticks as a container image for zero-installation execution and a Spack package for managed builds. The framework also incorporates safeguards against common pitfalls, such as deeply nested boolean operations in detector geometry, analyzing constructive solid geometry upon conversion from input files and issuing warnings to guide users toward optimized configurations. 👉 More information 🗞 GPU acceleration of optical photon propagation in low photon yield applications: Opticks for the Electron Ion Collider 🧠 ArXiv: https://arxiv.org/abs/2512.06061 Tags: