# GPU Tracking in the COMET Phase-I Cylindrical Drift Chamber

## TL;DR

This paper presents a GPU-accelerated track finding method for the COMET Phase-I experiment, significantly reducing computational costs by parallelizing track propagation and using initial seed guesses, improving efficiency in complex electron trajectory reconstruction.

## Contribution

It introduces a GPU-based approach combining parallel track propagation and initial seed estimation to efficiently solve the hit-to-turn assignment problem in cylindrical drift chambers.

## Key findings

- GPU implementation speeds up track finding compared to CPU
- Hough transform provides effective initial seed guesses
- Method effectively handles multi-turn electron trajectories

## Abstract

The GPU-accelerated track finding method is investigated to track electrons from neutrinoless muon decay in the COMET Phase-I experiment. Inside the cylindrical drift chamber, one third of the signal electron trajectories are composed of multiple turns where the correct hit assignments to each turn partition are significant in the track finding. Scanning all possible track seeds of position and momentum can resolve the hit-to-turn assignment problem with a high robustness, but requires a huge computational cost: The initial track seeds $(\theta,z,p_x,p_y,p_z)$ have broad uncertainties, so there exists many number of seeds that should be compared. In this article, this problem of massive computations are mitigated with 1) the parallel computing of Runge-Kutta-Nystr\"om track propagation with the GPU, and 2) an initial guess on the seeds using the Hough transform and the detector geometry. The computation speed enhancement compared to the CPU is also presented.