Testbed for GPU-accelerated applications 

Project goal

The goal of this project is to adapt computing models and software to exploit fully the potential of GPUs. The project, which began in late 2018, consists of ten individual use cases.

The technical coordinators are as follows:

Andrea Bocci, Felice Pantaleo, Maurizio Pierini, Federico Carminati, Vincenzo Innocente, Marco Rovere, Jean-Roch Vlimant, Vladimir Gligorov, Daniel Campora, Riccardo De Maria, Adrian Oeftiger, Lotta Mether, Ian Fisk, Lorenzo Moneta, Jan Kieseler, and Sofia Vallecorsa.

R&D topic
Computing performance and software
Project coordinator(s)
Maria Girone
Team members
Mary Touranakou, Thong Nguyen, Javier Duarte, Olmo Cerri, Roel Aaij, Dorothea Vom Bruch, Blaise Raheem Delaney, Ifan Williams, Niko Neufeld, Viktor Khristenko, Florian Reiss, Guillermo Izquierdo Moreno, Luca Atzori, Miguel Fontes Medeiros
Collaborator liaison(s)
Cosimo Gianfreda (E4), Daniele Gregori (E4), Agnese Reina (E4), Piero Altoé (NVIDIA), Andreas Hehn (NVIDIA), Tom Gibbs (NVIDIA)


Project background

Heterogeneous computing architectures will play an important role in helping CERN address the computing demands of the HL-LHC.

Recent progress

This CERN openlab project supports several use cases at CERN. This section outlines the progress made in the two main use cases that were worked on in 2020.

Allen: a high-level trigger on GPUs for LHCb

Allen’ is an initiative to develop a complete high-level trigger (the first step of the data-filtering process following particle collisions) on GPUs for the LHCb experiment. It has benefitted from support through CERN openlab, including consultation from engineers at NVIDIA.

The new system processes 40 Tb/s, using around 350 of the latest generation NVIDIA GPU cards. Allen matches — from a physics point of view — the reconstruction performance for charged particles achieved on traditional CPUs. It has also been shown that Allen will not be I/O or memory limited. Plus, not only can it be used to perform reconstruction, but it can also take decisions about whether to keep or reject events.

A diverse range of algorithms have been implemented efficiently on Allen. This demonstrates the potential for GPUs not only to be used as accelerators, but also as complete and standalone data-processing solutions.

In May 2020, Allen was adopted by the LHCb collaboration as the new baseline first-level trigger for Run 3. The Technical Design Report for the system was approved in June. From the start, Allen has been designed to be a framework that can be used in a general manner for high-throughput GPU computing. A workshop was held with core Gaudi developers and members of the CMS and ALICE  experiments to discuss how best to integrate Allen into the wider software ecosystem beyond the LHCb experiment. The LHCb team working on Allen is currently focusing on commissioning the system for data-taking in 2022 (delayed from the original 2021 start date due to the COVID-19 pandemic). A readiness review is taking place in the first half of 2021.

End-to-end multi-particle reconstruction for the HGCal based on machine learning

The CMS High-Granularity Calorimeter (HGCal) will replace the end-cap calorimeters of the CMS detector for the operation of the High-Luminosity LHC. With about 2 million sensors and high lateral and transversal granularity, it provides huge potential for new physics discoveries. We aim to exploit this using end-to-end optimisable graph neural networks.

Profiting from new machine-learning concepts developed within the group and the CERN openlab collaboration with the Flatiron Institute in New York, US, we were able to develop and train a first

prototype for directly reconstructing incident particle properties from raw detector hits. Through our direct contact with NVIDIA, we were able to implement custom tensorflow GPU kernels. Together with the dedicated neural network structure, these enabled us to process the hits of an entire particle-collision event in one go on the GPU.

Next steps

Work related to each of the project’s use cases will continue in 2021.


    A. Bocci, Towards a heterogeneous High Level Trigger farm for CMS (13 March). Presented at ACAT2019, Saas Fee, 2019. cern.ch/go/D9SF
    F. Pantaleo, Patatrack: accelerated Pixel Track reconstruction in CMS (2 April). Presented at Connecting the Dots 2019, Valencia, 2019. cern.ch/go/7D8W
    R. Kansal, Deep Graph Neural Networks for Fast HGCAL Simulation (13 August). Presented at CERN openlab summer-student lightning talk session, Geneva, 2019. cern.ch/go/qh6G
    A. Bocci, Heterogeneous reconstruction: combining an ARM processor with a GPU (4 November). Presented at CHEP2019, Adelaide, 2019. cern.ch/go/7bmH
    A. Bocci, Heterogeneous online reconstruction at CMS (7 November). Presented at 24th International Conference on Computing in High-Energy and Nuclear Physics (CHEP) 2019, Adelaide, 2019. cern.ch/go/l9JN