Efficiency studies for the gFEX hardware trigger in Phase I ATLAS upgrades

The global Feature Extractor (gFEX) is a Level 1 jet trigger that will be installed in ATLAS during Phase 1 upgrades. The gFEX will use large-radius jet algorithms refined with subjet information to select Lorentz-boosted objects like bosons and top quarks. With higher luminosities of the LHC approaching, triggers must cope with intense environments. Knowledge of the gFEX’s efficiency is critical as it is directly related to what data is captured by ATLAS. This study of gFEX efficiencies determines how well it performs on complicated objects and sets benchmarks for its performance using simulated samples assuming pp collisions at √s = 13 TeV with the ATLAS detector. The gFEX board includes a System on Chip with a GPU that is not currently utilized. These studies looked at the feasibility of utilizing the GPU for image classification or machine learning algorithms. I obtained metrics which provide important information on the GPU’s ability to process data. Further stress testing is needed to determine the true limits of the GPU function, but it looks probable that the GPU could be useful in the context of the gFEX hardware trigger to apply machine learning algorithms to generate further information on jets with complicated substructure.