Abstract
The upgraded CERN LHCb detector, due to start data taking in 2021, will have to reconstruct 4 TB/s of raw detector data in real time using commodity processors. This is one of the biggest real-time data processing challenges in any scientific domain. We present an intrinsically parallel reconstruction algorithm for the vertex detector of the LHCb experiment designed to optimally exploit multi-core general purpose architectures. We evaluate the algorithm on two high-end architectures from two different vendors and discuss in detail the impact of different SIMD Instruction Set Architecture extensions on the performance. We further compare the algorithm to current state-of-the-art scalar pattern recognition algorithms. We show a factor 2 speedup while achieving similar or better levels of physics performance.
Highlights
Components of a tracking algorithmA “hit” corresponds to an energy deposit in a single physical detector element
We present an efficient and highly parallel CPU implementation of one of the key LHCb track reconstruction algorithms, we show that it fits within the available resources for the LHCb upgrade, and we discuss the scaling of its performance with some Single Instruction Multiple Data (SIMD) Instruction Set Architecture (ISA) extensions on Intel’s Skylake and AMD’s Zen2 architectures
We have presented a new tracking algorithm for the VELO detector of the LHCb experiment specialized to take advantage of SIMD general purpose multicore processors
Summary
A “hit” corresponds to an energy deposit in a single physical detector element. Particle may leave multiple contiguous hits when traversing a single detector element. In this case these contiguous hits are grouped into “clusters” and the clusters are given as input to the tracking algorithm. A typical tracking algorithm consists of three logical elements: clustering, pattern recognition, and track fitting. Pattern recognition consists of choosing a subset of hits which correspond to a single particle traversing a detector. The quality of the track fit can be used to discriminate between genuine and fake tracks For this reason many tracking algorithms perform a partial fit during the pattern recognition stage and use its quality to reject the worst track candidates as early as possible. This metric should be as high as possible and is a good indicator of the quality of the reconstructed tracks.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
