Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 94: DAQ, Exp. Methods

T 94.4: Talk

Wednesday, March 22, 2023, 18:15–18:30, POT/0106

The Heterogeneous TICL Framework — •Wahid Redjeb^1,2, Abhirikshma Nandi¹, Alexander Schmidt¹, Felice Pantaleo², Marco Rovere², and Antonio Di Pilato³ — ¹III. Physikalisches Institut A, RWTH Aachen University, Aachen, Germany — ²CERN, Geneva, Switzerland — ³University of Geneva, Geneva, Switzerland

The High-Granularity Calorimeter (HGCAL) is a sampling calorimeter with both lateral and longitudinal fine granularity designed for the High-Luminosity LHC. The calorimeter will use silicon sensors, in the high radiation regions, providing high pile-up mitigation, and scintillators in the low radiations regions. For the physics object reconstruction a dedicated framework for HGCAL is currently under development: The Iterative Clustering (TICL), which utilizes the 5D (x,y,z,t,E) information from the reconstructed hits and returns particle properties and probabilities. Heterogeneous computing will play a fundamental role in the physics object reconstruction software to fully exploit the reach of the HL-LHC. Performance Portability libraries allow performance portability across different hardware architectures with a single code basis. In this talk we present and overview of the TICL framework and we show how heterogeneous computing has been integrated in the framework, exploiting the Alpaka library to achieve Performance Portability and being able to run core parts of the Framework on GPU and on CPU with a single source code.