Allen: A High-Level Trigger on GPUs for LHCb

Area of Science:

• High-energy physics
• Detector instrumentation
• Parallel computing

Background:

• The Large Hadron Collider beauty (LHCb) experiment requires a high-throughput first-level trigger for its 2021 upgrade.
• Existing trigger systems face challenges in processing the massive data rates generated by modern particle physics experiments.

Purpose of the Study:

• To develop and validate a fully Graphics Processing Unit (GPU)-based implementation for the LHCb upgrade's first-level trigger.
• To demonstrate the system's capability in handling high data rates and performing complex pattern recognition tasks.

Main Methods:

• A novel software implementation named Allen was developed, leveraging the parallel processing power of GPUs.
• The system was designed to perform tasks including charged particle trajectory finding, collision point reconstruction, particle identification (hadrons/muons), and displaced vertex finding.

Main Results:

• Allen successfully processes the 40 Tbit/s data rate of the upgraded LHCb detector.
• The system performs a wide range of pattern recognition tasks essential for physics analysis.
• Allen can be implemented on approximately 500 GPU cards and operates at the full LHC collision rate of 30 MHz without being I/O bound.

Conclusions:

• Allen represents the first complete high-throughput GPU trigger system proposed for a High Energy Physics (HEP) experiment.
• This GPU-based approach offers a scalable and efficient solution for real-time data processing in next-generation particle physics detectors.