A Comparison of CPU and GPU Implementations for the LHCb Experiment Run 3 Trigger

Aaij, R.; Adinolfi, M.; Aiola, S.; Akar, S.; Albrecht, J.; Alexander, M.; Amato, S.; Amhis, Y.; Archilli, F.; Bala, M Ponni; Bassi, G.; Bian, L.; Blago, M. P.; Boettcher, T.; Boldyrev, A. S.; Borghi, S.; Rodríguez, A. Brea; Calefice, Lukas; Gomez, M. Calvo; Pérez, Daniel Hugo Cámpora; Cardini, A.; Cattaneo, M.; Chobanova, V.; Ciezarek, G.; Vidal, X. Cid; Cobbledick, J. L.; Coelho, J. A. B.; Colombo, T.; Contu, A.; Couturier, B.; Craik, D. C.; Currie, R.; d’Argent, P.; Cian, M. De; Деркач, Д.; Dordei, F.; Dorigo, M.; Dufour, L.; Durante, P.; Dziurda, A.; Dzyuba, A.; Easo, S.; Esen, S.; Declara, P. Fernandez; Filippov, S.; Fitzpatrick, C.; Frank, M.; Gandini, P.; Gligorov, V. V.; Golobardes, Elisabet; Graziani, G.; Grillo, L.; Günther, P. A.; Hansmann-Menzemer, S.; Hennequin, A. M.; Henry, L.; Hill, D.; Hollitt, S. E.; Hu, J.; Hulsbergen, W.; Hunter, R. J.; Hushchyn, M.; Jashal, B. K.; Jones, C. R.; Klaver, S.; Klimaszewski, K.; Kopecna, R.; Krzemień, W.; Kucharczyk, M.; Lane, R.; Lazzari, F.; Gac, R. Le; Li, P.; Lopes, J. H.; Martínez, M. Lucio; Lupato, A.; Lupton, O.; Lyu, X. R.; Machefert, F.; Madejczyk, Olga; Malde, S.; Marchand, J. F.; Mariani, S.; Benito, C. Marín; Santos, D. Martínez; Vidal, F. Martínez; Matev, R.; Mazurek, M.; Mitreska, B.; Mitzel, D. S.; Morello, M. J.; Mu, H.; Muzzetto, P.; Naik, P.; Needham, M.; Neri, N.; Neufeld, N.; Nolte, N. S.; O’Hanlon, D. P.; Oyanguren, A.; Altarelli, M. Pepe; Petrucci, S.; Petruzzo, M.; Pica, Lorenzo; Pisani, F.; Piucci, A.; Polci, F.; Poluektov, A.; Polycarpo, E.; Prouvé, C.; Punzi, G.; Quagliani, R.; Trejo, Raul Iraq Rabadan; Pernas, M. Ramos; Rangel, M. S.; Ratnikov, F.; Raven, G.; Reiss, F.; Renaudin, V.; Robbe, P.; Ryzhikov, A.; Santimaria, M.; Saur, M.; Schiller, M.; Schwemmer, R.; Sciascia, B.; Solomin, A.; Suljik, F.; Skidmore, N.; Sokoloff, M. D.; Spradlin, P.; Stahl, M.; Stahl, S.; Stevens, H.; Sun, L.; Szabelski, A.; Szumlak, T.; Szymanski, Marc; Tou, D. Y.; Tuci, G.; Usachov, A.; Canudas, N. Valls; Gomez, R. Vazquez; Vecchi, S.; Vesterinen, M.; Vilasís-Cardona, X.; Bruch, Dorothea Vom; Wang, Z.; Wojtoń, Tomasz; Whitehead, M.; Williams, M.; Witek, M.; Xie, Y.; Xu, Ao; Yin, H.; Zdybał, M.; Zenaiev, O.; Zhang, D.; Zhang, L.; Zhu, X.

doi:10.1007/s41781-021-00070-2

Cited by 12 publications

(6 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The two configurations were tested on both CPU and GPU architectures, demonstrating compatible results in terms of event rates. A more detailed comparison of the CPU to GPU physics performance of Allen based on simulation can be found in [8].…”

Section: Integration Activitiesmentioning

confidence: 99%

Commissioning LHCb’s GPU high level trigger

Agapopoulou

2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

From 2022 onward, the upgraded LHCb experiment will use a triggerless readout system collecting data at an event rate of 30 MHz. A software-only High Level Trigger will enable unprecedented flexibility for trigger selections. During the first stage (HLT1), a sub-set of the full offline track reconstruction for charged particles is run to select particles of interest based on single or two-track selections. This includes decoding the raw data, clustering of hits, pattern recognition, as well as track fitting, ghost track rejection with machine learning techniques and finally event selections. HLT1 will be processed entirely on O(100) GPUs. We will present the integration of the GPU HLT1 into LHCb’s data acquisition system in preparation for data taking in 2022.

show abstract

Section: Integration Activitiesmentioning

confidence: 99%

Commissioning LHCb’s GPU high level trigger

Agapopoulou

2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

show abstract

“…In recent years, GPU acceleration has been investigated as a means to speed up traditional particle-reconstruction algorithms by parallelization [9][10][11][12][13][14][15]. In view of these promis-ing results, the LHC experimental collaborations invested financial resources to migrate their computing infrastructure towards CPU+GPU heterogeneous computing [16][17][18].…”

Section: Related Workmentioning

confidence: 99%

End-to-end multi-particle reconstruction in high occupancy imaging calorimeters with graph neural networks

et al. 2022

View full text Add to dashboard Cite

We present an end-to-end reconstruction algorithm to build particle candidates from detector hits in next-generation granular calorimeters similar to that foreseen for the high-luminosity upgrade of the CMS detector. The algorithm exploits a distance-weighted graph neural network, trained with object condensation, a graph segmentation technique. Through a single-shot approach, the reconstruction task is paired with energy regression. We describe the reconstruction performance in terms of efficiency as well as in terms of energy resolution. In addition, we show the jet reconstruction performance of our method and discuss its inference computational cost. To our knowledge, this work is the first-ever example of single-shot calorimetric reconstruction of $${\mathcal {O}}(1000)$$ O ( 1000 ) particles in high-luminosity conditions with 200 pileup.

show abstract

“…In recent years, GPU acceleration has been investigated as a means to speed up traditional particle-reconstruction algorithms by parallelization [9][10][11][12][13][14][15]. In view of these promising results, the LHC experimental collaborations invested financial resources to migrate their computing infrastructure towards CPU+GPU heterogeneous computing [16][17][18].…”

Section: Related Workmentioning

confidence: 99%

End-to-end multi-particle reconstruction in high occupancy imaging calorimeters with graph neural networks

Qasim,

Chernyavskaya,

Kieseler

et al. 2022

Preprint

View full text Add to dashboard Cite

We present an end-to-end reconstruction algorithm to build particle candidates from detector hits in nextgeneration granular calorimeters similar to that foreseen for the high-luminosity upgrade of the CMS detector. The algorithm exploits a distance-weighted graph neural network, trained with object condensation, a graph segmentation technique. Through a single-shot approach, the reconstruction task is paired with energy regression. We describe the reconstruction performance in terms of efficiency as well as in terms of energy resolution. In addition, we show the jet reconstruction performance of our method and discuss its inference computational cost. To our knowledge, this work is the first-ever example of single-shot calorimetric reconstruction of O(1000) particles in high-luminosity conditions with 200 pileup.

show abstract

A Comparison of CPU and GPU Implementations for the LHCb Experiment Run 3 Trigger

Cited by 12 publications

References 6 publications

Commissioning LHCb’s GPU high level trigger

Commissioning LHCb’s GPU high level trigger

End-to-end multi-particle reconstruction in high occupancy imaging calorimeters with graph neural networks

End-to-end multi-particle reconstruction in high occupancy imaging calorimeters with graph neural networks

Contact Info

Product

Resources

About