SMuK 2023 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 127: Exp. Methods III

T 127.3: Vortrag

Donnerstag, 23. März 2023, 16:20–16:35, WIL/C129

Navigation and track parameter transport using a heterogeneous code design for CPUs and GPUs within the ACTS R&D project — Andreas Salzburger¹, •Joana Niermann^{1, 2}, Beomki Yeo^{3, 4}, Stephen Swatman^{1, 5}, Attila Krasznahorkay¹, and Stan Lai² — ¹CERN — ²II. Physikalisches Institut, Georg-August-Universität Göttingen — ³Department of Physics, University of California — ⁴Lawrence Berkeley National Laboratory — ⁵University of Amsterdam

With the upcoming high luminosity era of the LHC, track reconstruction, in particular, will suffer from drastically increasing combinatorics. A promising perspective to meet these rising computing demands is the deployment of hardware accelerators which offer massive parallelism, like GPGPUs. Current state-of-the-art implementations of pattern recognition algorithms in track reconstruction are problematic to adapt to accelerator hardware architectures in several ways. For example, runtime-polymorphic geometry classes and pointer based data structures are commonly difficult to move to an accelerator device.

Within the ACTS parallelization R&D project, research is on-going to adapt a complete track reconstruction chain, from clusterization to track fitting, to run efficiently on GPUs. We show the implementation and performance of a core component of this chain: the propagation of track parameters and their associated covariances through an inhomogeneous magnetic field (covfie library), together with the application of material effects. The implementation is part of the detray library and makes use of its geometry description and navigation.