A parallel algorithm for fast reconstruction of primary vertices on heterogeneous architectures

TL;DR

This paper presents a new parallel algorithm for rapid primary vertex reconstruction in high-energy physics experiments, optimized for heterogeneous architectures like CPUs and GPUs, enabling efficient online processing at the LHCb experiment.

Contribution

It introduces a novel cluster-based vertex reconstruction algorithm tailored for heterogeneous architectures, improving speed and efficiency for real-time particle collision analysis.

Findings

Achieves fast vertex reconstruction suitable for online trigger systems

Demonstrates effective implementation on x86 and GPU architectures

Shows promising performance on simulated collision data

Abstract

The physics programme of the LHCb experiment at the Large Hadron Collider requires an efficient and precise reconstruction of the particle collision vertices. The LHCb Upgrade detector relies on a fully software-based trigger with an online reconstruction rate of 30 MHz, necessitating fast vertex finding algorithms. This paper describes a new approach to vertex reconstruction developed for this purpose. The algorithm is based on cluster finding within a histogram of the particle trajectory projections along the beamline and on an adaptive vertex fit. Its implementations and optimisations on x86 and GPU architectures and its performance on simulated samples are also discussed.