Göttingen 2025 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 55: Data, AI, Computing, Electronics VI (DAQ and Trigger)
T 55.7: Vortrag
Mittwoch, 2. April 2025, 17:45–18:00, VG 2.102
A parametrised Kalman filter for the GPU-based first level trigger of the upgraded LHCb experiment — Michel De Cian1,2, Stephanie Hansmann-Menzemer1, and •Lennart Uecker1 — 1Physikalisches Institut, Ruprecht-Karls-Universität Heidelberg, Germany — 2Department of Physics and Astronomy, University of Manchester, United Kingdom
The LHCb Upgrade I detector implements a fully software-based trigger to select a wide range of physics signatures. The first-level trigger employs 500 GPU cards to perform partial event reconstruction at the complete LHCb collision rate of 30 MHz. This includes finding charged tracks, reconstructing proton-proton collisions, particle identification and finding displaced vertices. The implementation enables an increased trigger performance, especially for hadronic channels, thus elevating signal yields above the increase in luminosity.
Kalman filters are commonly used in High Energy Physics to process pattern recognition tracks and extract optimal track parameters. However, these filters' sequential nature and substantial memory requirements make them suboptimal for GPU implementation. We developed a method that replaces computationally demanding operations with parametrized approximations, specifically material scattering calculations and state extrapolation in the magnetic field. This approach achieves high-throughput track fitting while maintaining track information quality, making it suitable for GPU-based processing in high-rate environments. In this talk we present the implementation and performance of the GPU based Kalman filter for LHCb Run 3.
Keywords: Trigger; GPU; Kalman Filter; LHCb