Speaker
Description
ALICE is the dedicated heavy-ion experiment at the LHC at CERN.
For LHC Run 3, ALICE upgraded the TPC with a GEM readout to support continuous operation and the recording of all Pb-Pb collisions.
The peak Pb-Pb recording rate increased from around 1 kHz to 50 kHz.
In order to keep up with the increased data rate, ALICE upgraded its whole online computing scheme, and performs full online processing of all recorded data for calibration and for data compression.
ALICE was the first LHC experiment to leverage GPUs for online data processing during LHC Runs 1 and 2.
In Run 3, TPC processing accounts for more than 90% of the total online workload and runs entirely on GPUs.
This includes clusterization, tracking, and compression.
To efficiently use its online processing farm for offline reconstruction when the LHC is not operating, ALICE has been running offline TPC tracking on GPUs since 2023.
Meanwhile, ALICE has offloaded additional reconstruction steps to GPUs to improve the utilization of GPUs during offline reconstruction.
Online track reconstruction of a TPC with continuous readout posed several new challenges, in particular the absense of a priori knowledge of the z position of hits in the TPC.
The talk will give an overview of the novelties of TPC reconstruction in Run 3, and in particular of the challenges of online processing in such high-rate environments.
