Real-time data selection on GPUs at the LHCb experiment

by Dorothea Vom Bruch (CPPM, Marseille)

Friday 20 Nov 2020, 12:30 → 14:00 Europe/London

https://zoom.us/j/304578256

The LHCb experiment at CERN – designed to study the properties of beauty and charm hadrons – is currently undergoing a major upgrade with a replacement of almost all components. The new detector will start data taking in Run 3 of the LHC, to begin in 2022.

One key feature of the upgrade is the readout of all sub-detectors at the proton bunch crossing rate of 30 MHz, corresponding to a data rate of up to 40 Tbit/s. Data reconstruction and selection is done entirely in software, allowing for a high degree of flexibility. This results in a significant increase in selection efficiency compared to the old LHCb readout system, where proton bunch crossings were first analyzed by algorithms implemented in hardware and only selected bunch crossings were read out entirely.

For an efficient first stage of data reduction the trajectories of charged particles in the detector have to be reconstructed in real-time to select proton bunch collisions of interest.
This poses a huge computing challenge at 30 MHz.
In the seminar, I will discuss how we make use of the compute power of graphics processing units (GPUs) to process the entire first selection stage (High Level Trigger 1).

Connections will begin at 12:30 with the seminar beginning at 13:00.

vomBruch_Niels_Bohr_seminar_LHCb_trigger_GPUs_11-2020.pdf