traccc: GPU track reconstruction library for HEP experiments

TL;DR

traccc is a GPU-based track reconstruction library for high energy physics experiments, demonstrating comparable physics accuracy to CPU methods and significantly improved computational performance on GPUs for large datasets.

Contribution

This work introduces traccc, a GPU-accelerated library implementing HEP tracking algorithms, with benchmarking showing its efficiency and accuracy relative to CPU-based methods.

Findings

GPU implementation achieves similar physics performance as CPU

GPUs outperform CPUs in computational speed for large events

traccc advances GPU utilization in HEP data processing

Abstract

We present the current development status and progress of traccc, a GPU track reconstruction library developed in the context of the A Common Tracking Software (ACTS) project. traccc implements tracking algorithms used in high energy physics (HEP) experiments, including the Kalman filter based track finding and fitting. We benchmark the software with data simulated by Geant4 to measure the physics and computing performance. We show that the physics performance for GPU and CPU are very close. We also show that the GPUs can achieve higher computational performance than the CPU for sufficiently large events.