Speaker
Description
The planned upgrades to the Large Hadron Collider for the HL-LHC era will gradually increase the nominal luminosity, aiming for peak values close to 5×10$^{34}$ cm$^{-2}$s$^{-1}$ in the ATLAS and CMS experiments. With higher luminosity, more proton–proton interactions will happen in each bunch crossing, with pileup reaching up to 200, making track reconstruction much more challenging.
To handle these conditions, several experiments have started updating parts of their track reconstruction software so it can run efficiently on different types of computing hardware. These efforts have shown good progress but usually stay within each experiment’s own software.
In this work, we present a standalone framework that can run on multiple backends—including CPUs, NVIDIA GPUs, and AMD GPUs, and reconstruct tracks in cylindrical tracking detectors used by various high-energy physics experiments. We evaluate both its physics performance and its computational behavior for different detector setups.
This project is an initial step toward a unified, experiment-independent reconstruction tool for HL-LHC-like detectors, able to use heterogeneous computing resources. The detector model is defined only by its usual components: a silicon tracking system, at least one calorimeter, and a muon subsystem.
| I read the instructions above | Yes |
|---|