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

The LHCb Collaboration

doi:10.1007/s41781-021-00070-2

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

The LHCb Collaboration

(Astro)-Particles Physics

Research output: Contribution to Journal › Article › Academic › peer-review

Abstract

The Large Hadron Collider beauty (LHCb) experiment at CERN is undergoing an upgrade in preparation for the Run 3 data collection period at the Large Hadron Collider (LHC). As part of this upgrade, the trigger is moving to a full software implementation operating at the LHC bunch crossing rate. We present an evaluation of a CPU-based and a GPU-based implementation of the first stage of the high-level trigger. After a detailed comparison, both options are found to be viable. This document summarizes the performance and implementation details of these options, the outcome of which has led to the choice of the GPU-based implementation as the baseline.

Original language	English
Article number	1
Pages (from-to)	1-20
Number of pages	20
Journal	Computing and Software for Big Science
Volume	6
Issue number	1
Early online date	22 Dec 2022
DOIs	https://doi.org/10.1007/s41781-021-00070-2
Publication status	Published - Dec 2022

Bibliographical note

Funding Information:
We thank the technical and administrative staff at the LHCb institutes. We acknowledge support from CERN and from the national agencies: CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and NSFC (China); CNRS/IN2P3 (France); BMBF, DFG and MPG (Germany); INFN (Italy); NWO (Netherlands); MEiN and NCN UMO-2018/31/B/ST2/03998 (Poland); MEN/IFA (Romania); MSHE (Russia); MICINN (Spain); SNSF and SER (Switzerland); NASU (Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the computing resources that are provided by CERN, IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), PIC (Spain), GridPP (United Kingdom), RRCKI and Yandex LLC (Russia), CSCS (Switzerland), IFIN-HH (Romania), CBPF (Brazil), PL-GRID (Poland) and NERSC (USA). We are indebted to the communities behind the multiple open-source software packages on which we depend. Individual groups or members have received support from ARC and ARDC (Australia); AvH Foundation (Germany); EPLANET, Marie Skłodowska-Curie Actions and ERC (European Union); A*MIDEX, ANR, IPhU and Labex P2IO, and Région Auvergne-Rhône-Alpes (France); Key Research Program of Frontier Sciences of CAS, CAS PIFI, CAS CCEPP, Fundamental Research Funds for the Central Universities, and Sci. & Tech. Program of Guangzhou (China); RFBR, RSF and Yandex LLC (Russia); GVA, XuntaGal and GENCAT (Spain); the Leverhulme Trust, the Royal Society and UKRI (United Kingdom). The authors would like to thank the LHCb computing and simulation teams for their support and for producing the simulated LHCb samples used to benchmark the performance of RTA software.

Funding Information:
We thank the technical and administrative staff at the LHCb institutes. We acknowledge support from CERN and from the national agencies: CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and NSFC (China); CNRS/IN2P3 (France); BMBF, DFG and MPG (Germany); INFN (Italy); NWO (Netherlands); MEiN and NCN UMO-2018/31/B/ST2/03998 (Poland); MEN/IFA (Romania); MSHE (Russia); MICINN (Spain); SNSF and SER (Switzerland); NASU (Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the computing resources that are provided by CERN, IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), PIC (Spain), GridPP (United Kingdom), RRCKI and Yandex LLC (Russia), CSCS (Switzerland), IFIN-HH (Romania), CBPF (Brazil), PL-GRID (Poland) and NERSC (USA). We are indebted to the communities behind the multiple open-source software packages on which we depend. Individual groups or members have received support from ARC and ARDC (Australia); AvH Foundation (Germany); EPLANET, Marie Sk?odowska-Curie Actions and ERC (European Union); A*MIDEX, ANR, IPhU and Labex P2IO, and R?gion Auvergne-Rh?ne-Alpes (France); Key Research Program of Frontier Sciences of CAS, CAS PIFI, CAS CCEPP, Fundamental Research Funds for the Central Universities, and Sci. & Tech. Program of Guangzhou (China); RFBR, RSF and Yandex LLC (Russia); GVA, XuntaGal and GENCAT (Spain); the Leverhulme Trust, the Royal Society and UKRI (United Kingdom). The authors would like to thank the LHCb computing and simulation teams for their support and for producing the simulated LHCb samples used to benchmark the performance of RTA software.

Publisher Copyright:
© 2021, The Author(s).

Funding

We thank the technical and administrative staff at the LHCb institutes. We acknowledge support from CERN and from the national agencies: CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and NSFC (China); CNRS/IN2P3 (France); BMBF, DFG and MPG (Germany); INFN (Italy); NWO (Netherlands); MEiN and NCN UMO-2018/31/B/ST2/03998 (Poland); MEN/IFA (Romania); MSHE (Russia); MICINN (Spain); SNSF and SER (Switzerland); NASU (Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the computing resources that are provided by CERN, IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), PIC (Spain), GridPP (United Kingdom), RRCKI and Yandex LLC (Russia), CSCS (Switzerland), IFIN-HH (Romania), CBPF (Brazil), PL-GRID (Poland) and NERSC (USA). We are indebted to the communities behind the multiple open-source software packages on which we depend. Individual groups or members have received support from ARC and ARDC (Australia); AvH Foundation (Germany); EPLANET, Marie Skłodowska-Curie Actions and ERC (European Union); A*MIDEX, ANR, IPhU and Labex P2IO, and Région Auvergne-Rhône-Alpes (France); Key Research Program of Frontier Sciences of CAS, CAS PIFI, CAS CCEPP, Fundamental Research Funds for the Central Universities, and Sci. & Tech. Program of Guangzhou (China); RFBR, RSF and Yandex LLC (Russia); GVA, XuntaGal and GENCAT (Spain); the Leverhulme Trust, the Royal Society and UKRI (United Kingdom). The authors would like to thank the LHCb computing and simulation teams for their support and for producing the simulated LHCb samples used to benchmark the performance of RTA software. We thank the technical and administrative staff at the LHCb institutes. We acknowledge support from CERN and from the national agencies: CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and NSFC (China); CNRS/IN2P3 (France); BMBF, DFG and MPG (Germany); INFN (Italy); NWO (Netherlands); MEiN and NCN UMO-2018/31/B/ST2/03998 (Poland); MEN/IFA (Romania); MSHE (Russia); MICINN (Spain); SNSF and SER (Switzerland); NASU (Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the computing resources that are provided by CERN, IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), PIC (Spain), GridPP (United Kingdom), RRCKI and Yandex LLC (Russia), CSCS (Switzerland), IFIN-HH (Romania), CBPF (Brazil), PL-GRID (Poland) and NERSC (USA). We are indebted to the communities behind the multiple open-source software packages on which we depend. Individual groups or members have received support from ARC and ARDC (Australia); AvH Foundation (Germany); EPLANET, Marie Sk?odowska-Curie Actions and ERC (European Union); A*MIDEX, ANR, IPhU and Labex P2IO, and R?gion Auvergne-Rh?ne-Alpes (France); Key Research Program of Frontier Sciences of CAS, CAS PIFI, CAS CCEPP, Fundamental Research Funds for the Central Universities, and Sci. & Tech. Program of Guangzhou (China); RFBR, RSF and Yandex LLC (Russia); GVA, XuntaGal and GENCAT (Spain); the Leverhulme Trust, the Royal Society and UKRI (United Kingdom). The authors would like to thank the LHCb computing and simulation teams for their support and for producing the simulated LHCb samples used to benchmark the performance of RTA software.

Funders	Funder number
ARDC
CAS CCEPP
EPLANET
GridPP
IFIN-HH
Key Research Program of Frontier Sciences of CAS
MSHE
PL-GRID
Program of Guangzhou
RRCKI
SURF
XuntaGal
Yandex LLC
National Science Foundation
U.S. Department of Energy
Alexander von Humboldt-Stiftung
Association pour la Recherche sur le Cancer
Automotive Research Center
Karlsruhe Institute of Technology
H2020 Marie Skłodowska-Curie Actions
CERN
UK Research and Innovation
National Energy Research Scientific Computing Center
Science and Technology Facilities Council
Leverhulme Trust
Royal Society
European Research Council
Deutsches Elektronen-Synchrotron
Deutsche Forschungsgemeinschaft
Agence Nationale de la Recherche
Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung
National Natural Science Foundation of China
Russian Foundation for Basic Research
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
Bundesministerium für Bildung und Forschung
Chinese Academy of Sciences
Nederlandse Organisatie voor Wetenschappelijk Onderzoek
Generalitat Valenciana
Conselho Nacional de Desenvolvimento Científico e Tecnológico
Instituto Nazionale di Fisica Nucleare
Narodowe Centrum Nauki	UMO-2018/31/B/ST2/03998
Narodowe Centrum Nauki
Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro
Ministry of Science and Technology, Taiwan
National Academy of Sciences of Ukraine
Centre National de la Recherche Scientifique
Financiadora de Estudos e Projetos
Ministerio de Ciencia e Innovación
Russian Science Foundation
Région Auvergne-Rhône-Alpes
Fundamental Research Funds for the Central Universities
Institut national de physique nucléaire et de physique des particules
Sociedad Española de Reumatología

Keywords

Heterogeneous
High-level trigger
High-throughput
Parallel computing
Real-time
Software

Access to Document

10.1007/s41781-021-00070-2

Persistent URL (handle)

Persistent link

Cite this

@article{8615ed23ce53425095398c03aadceb98,

title = "A Comparison of CPU and GPU Implementations for the LHCb Experiment Run 3 Trigger",

abstract = "The Large Hadron Collider beauty (LHCb) experiment at CERN is undergoing an upgrade in preparation for the Run 3 data collection period at the Large Hadron Collider (LHC). As part of this upgrade, the trigger is moving to a full software implementation operating at the LHC bunch crossing rate. We present an evaluation of a CPU-based and a GPU-based implementation of the first stage of the high-level trigger. After a detailed comparison, both options are found to be viable. This document summarizes the performance and implementation details of these options, the outcome of which has led to the choice of the GPU-based implementation as the baseline.",

keywords = "Heterogeneous, High-level trigger, High-throughput, Parallel computing, Real-time, Software",

author = "R. Aaij and M. Adinolfi and S. Aiola and S. Akar and J. Albrecht and M. Alexander and S. Amato and Y. Amhis and F. Archilli and M. Bala and G. Bassi and L. Bian and Blago, {M. P.} and T. Boettcher and A. Boldyrev and S. Borghi and Rodriguez, {A. Brea} and L. Calefice and Gomez, {M. Calvo} and P{\'e}rez, {D. H.C{\'a}mpora} and A. Cardini and M. Cattaneo and V. Chobanova and G. Ciezarek and Vidal, {X. Cid} and Cobbledick, {J. L.} and Coelho, {J. A.B.} and T. Colombo and A. Contu and B. Couturier and Craik, {D. C.} and R. Currie and P. d{\textquoteright}Argent and {De Cian}, M. and D. Derkach and F. Dordei and M. Dorigo and L. Dufour and P. Durante and A. Dziurda and A. Dzyuba and S. Easo and S. Esen and Declara, {P. Fernandez} and S. Filippov and C. Fitzpatrick and M. Frank and P. Gandini and Gligorov, {V. V.} and E. Golobardes and G. Graziani and L. Grillo and G{\"u}nther, {P. A.} and S. Hansmann-Menzemer and Hennequin, {A. M.} and L. Henry and D. Hill and Hollitt, {S. E.} and J. Hu and W. Hulsbergen and Hunter, {R. J.} and M. Hushchyn and Jashal, {B. K.} and Jones, {C. R.} and S. Klaver and K. Klimaszewski and R. Kopecna and W. Krzemien and M. Kucharczyk and R. Lane and F. Lazzari and Gac, {R. Le} and P. Li and Lopes, {J. H.} and Martinez, {M. Lucio} and A. Lupato and O. Lupton and X. Lyu and F. Machefert and O. Madejczyk and S. Malde and Marchand, {J. F.} and S. Mariani and Benito, {C. Marin} and Santos, {D. Martinez} and Vidal, {F. Martinez} and R. Matev and M. Mazurek and B. Mitreska and Mitzel, {D. S.} and Morello, {M. J.} and H. Mu and P. Muzzetto and P. Naik and M. Needham and N. Neri and N. Neufeld and Nolte, {N. S.} and D. O{\textquoteright}Hanlon and A. Oyanguren and Altarelli, {M. Pepe} and S. Petrucci and M. Petruzzo and L. Pica and F. Pisani and A. Piucci and F. Polci and A. Poluektov and E. Polycarpo and C. Prouve and G. Punzi and R. Quagliani and Trejo, {R. I.Rabadan} and Pernas, {M. Ramos} and Rangel, {M. S.} and F. Ratnikov and G. Raven and F. Reiss and V. Renaudin and P. Robbe and A. Ryzhikov and M. Santimaria and M. Saur and M. Schiller and R. Schwemmer and B. Sciascia and A. Solomin and F. Suljik and N. Skidmore and Sokoloff, {M. D.} and P. Spradlin and M. Stahl and S. Stahl and H. Stevens and L. Sun and A. Szabelski and T. Szumlak and M. Szymanski and Tou, {D. Y.} and G. Tuci and A. Usachov and Canudas, {N. Valls} and Gomez, {R. Vazquez} and S. Vecchi and M. Vesterinen and X. Vilasis-Cardona and Bruch, {D. Vom} and Z. Wang and T. Wojton and M. Whitehead and M. Williams and M. Witek and Y. Xie and A. Xu and H. Yin and M. Zdybal and O. Zenaiev and D. Zhang and L. Zhang and X. Zhu and {The LHCb Collaboration}",

note = "Funding Information: We thank the technical and administrative staff at the LHCb institutes. We acknowledge support from CERN and from the national agencies: CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and NSFC (China); CNRS/IN2P3 (France); BMBF, DFG and MPG (Germany); INFN (Italy); NWO (Netherlands); MEiN and NCN UMO-2018/31/B/ST2/03998 (Poland); MEN/IFA (Romania); MSHE (Russia); MICINN (Spain); SNSF and SER (Switzerland); NASU (Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the computing resources that are provided by CERN, IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), PIC (Spain), GridPP (United Kingdom), RRCKI and Yandex LLC (Russia), CSCS (Switzerland), IFIN-HH (Romania), CBPF (Brazil), PL-GRID (Poland) and NERSC (USA). We are indebted to the communities behind the multiple open-source software packages on which we depend. Individual groups or members have received support from ARC and ARDC (Australia); AvH Foundation (Germany); EPLANET, Marie Sk{\l}odowska-Curie Actions and ERC (European Union); A*MIDEX, ANR, IPhU and Labex P2IO, and R{\'e}gion Auvergne-Rh{\^o}ne-Alpes (France); Key Research Program of Frontier Sciences of CAS, CAS PIFI, CAS CCEPP, Fundamental Research Funds for the Central Universities, and Sci. & Tech. Program of Guangzhou (China); RFBR, RSF and Yandex LLC (Russia); GVA, XuntaGal and GENCAT (Spain); the Leverhulme Trust, the Royal Society and UKRI (United Kingdom). The authors would like to thank the LHCb computing and simulation teams for their support and for producing the simulated LHCb samples used to benchmark the performance of RTA software. Funding Information: We thank the technical and administrative staff at the LHCb institutes. We acknowledge support from CERN and from the national agencies: CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and NSFC (China); CNRS/IN2P3 (France); BMBF, DFG and MPG (Germany); INFN (Italy); NWO (Netherlands); MEiN and NCN UMO-2018/31/B/ST2/03998 (Poland); MEN/IFA (Romania); MSHE (Russia); MICINN (Spain); SNSF and SER (Switzerland); NASU (Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the computing resources that are provided by CERN, IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), PIC (Spain), GridPP (United Kingdom), RRCKI and Yandex LLC (Russia), CSCS (Switzerland), IFIN-HH (Romania), CBPF (Brazil), PL-GRID (Poland) and NERSC (USA). We are indebted to the communities behind the multiple open-source software packages on which we depend. Individual groups or members have received support from ARC and ARDC (Australia); AvH Foundation (Germany); EPLANET, Marie Sk?odowska-Curie Actions and ERC (European Union); A*MIDEX, ANR, IPhU and Labex P2IO, and R?gion Auvergne-Rh?ne-Alpes (France); Key Research Program of Frontier Sciences of CAS, CAS PIFI, CAS CCEPP, Fundamental Research Funds for the Central Universities, and Sci. & Tech. Program of Guangzhou (China); RFBR, RSF and Yandex LLC (Russia); GVA, XuntaGal and GENCAT (Spain); the Leverhulme Trust, the Royal Society and UKRI (United Kingdom). The authors would like to thank the LHCb computing and simulation teams for their support and for producing the simulated LHCb samples used to benchmark the performance of RTA software. Publisher Copyright: {\textcopyright} 2021, The Author(s).",

year = "2022",

month = dec,

doi = "10.1007/s41781-021-00070-2",

language = "English",

volume = "6",

pages = "1--20",

journal = "Computing and Software for Big Science",

issn = "2510-2044",

publisher = "Springer Nature",

number = "1",

}

TY - JOUR

T1 - A Comparison of CPU and GPU Implementations for the LHCb Experiment Run 3 Trigger

AU - Aaij, R.

AU - Adinolfi, M.

AU - Aiola, S.

AU - Akar, S.

AU - Albrecht, J.

AU - Alexander, M.

AU - Amato, S.

AU - Amhis, Y.

AU - Archilli, F.

AU - Bala, M.

AU - Bassi, G.

AU - Bian, L.

AU - Blago, M. P.

AU - Boettcher, T.

AU - Boldyrev, A.

AU - Borghi, S.

AU - Rodriguez, A. Brea

AU - Calefice, L.

AU - Gomez, M. Calvo

AU - Pérez, D. H.Cámpora

AU - Cardini, A.

AU - Cattaneo, M.

AU - Chobanova, V.

AU - Ciezarek, G.

AU - Vidal, X. Cid

AU - Cobbledick, J. L.

AU - Coelho, J. A.B.

AU - Colombo, T.

AU - Contu, A.

AU - Couturier, B.

AU - Craik, D. C.

AU - Currie, R.

AU - d’Argent, P.

AU - De Cian, M.

AU - Derkach, D.

AU - Dordei, F.

AU - Dorigo, M.

AU - Dufour, L.

AU - Durante, P.

AU - Dziurda, A.

AU - Dzyuba, A.

AU - Easo, S.

AU - Esen, S.

AU - Declara, P. Fernandez

AU - Filippov, S.

AU - Fitzpatrick, C.

AU - Frank, M.

AU - Gandini, P.

AU - Gligorov, V. V.

AU - Golobardes, E.

AU - Graziani, G.

AU - Grillo, L.

AU - Günther, P. A.

AU - Hansmann-Menzemer, S.

AU - Hennequin, A. M.

AU - Henry, L.

AU - Hill, D.

AU - Hollitt, S. E.

AU - Hu, J.

AU - Hulsbergen, W.

AU - Hunter, R. J.

AU - Hushchyn, M.

AU - Jashal, B. K.

AU - Jones, C. R.

AU - Klaver, S.

AU - Klimaszewski, K.

AU - Kopecna, R.

AU - Krzemien, W.

AU - Kucharczyk, M.

AU - Lane, R.

AU - Lazzari, F.

AU - Gac, R. Le

AU - Li, P.

AU - Lopes, J. H.

AU - Martinez, M. Lucio

AU - Lupato, A.

AU - Lupton, O.

AU - Lyu, X.

AU - Machefert, F.

AU - Madejczyk, O.

AU - Malde, S.

AU - Marchand, J. F.

AU - Mariani, S.

AU - Benito, C. Marin

AU - Santos, D. Martinez

AU - Vidal, F. Martinez

AU - Matev, R.

AU - Mazurek, M.

AU - Mitreska, B.

AU - Mitzel, D. S.

AU - Morello, M. J.

AU - Mu, H.

AU - Muzzetto, P.

AU - Naik, P.

AU - Needham, M.

AU - Neri, N.

AU - Neufeld, N.

AU - Nolte, N. S.

AU - O’Hanlon, D.

AU - Oyanguren, A.

AU - Altarelli, M. Pepe

AU - Petrucci, S.

AU - Petruzzo, M.

AU - Pica, L.

AU - Pisani, F.

AU - Piucci, A.

AU - Polci, F.

AU - Poluektov, A.

AU - Polycarpo, E.

AU - Prouve, C.

AU - Punzi, G.

AU - Quagliani, R.

AU - Trejo, R. I.Rabadan

AU - Pernas, M. Ramos

AU - Rangel, M. S.

AU - Ratnikov, F.

AU - Raven, G.

AU - Reiss, F.

AU - Renaudin, V.

AU - Robbe, P.

AU - Ryzhikov, A.

AU - Santimaria, M.

AU - Saur, M.

AU - Schiller, M.

AU - Schwemmer, R.

AU - Sciascia, B.

AU - Solomin, A.

AU - Suljik, F.

AU - Skidmore, N.

AU - Sokoloff, M. D.

AU - Spradlin, P.

AU - Stahl, M.

AU - Stahl, S.

AU - Stevens, H.

AU - Sun, L.

AU - Szabelski, A.

AU - Szumlak, T.

AU - Szymanski, M.

AU - Tou, D. Y.

AU - Tuci, G.

AU - Usachov, A.

AU - Canudas, N. Valls

AU - Gomez, R. Vazquez

AU - Vecchi, S.

AU - Vesterinen, M.

AU - Vilasis-Cardona, X.

AU - Bruch, D. Vom

AU - Wang, Z.

AU - Wojton, T.

AU - Whitehead, M.

AU - Williams, M.

AU - Witek, M.

AU - Xie, Y.

AU - Xu, A.

AU - Yin, H.

AU - Zdybal, M.

AU - Zenaiev, O.

AU - Zhang, D.

AU - Zhang, L.

AU - Zhu, X.

AU - The LHCb Collaboration

N1 - Funding Information: We thank the technical and administrative staff at the LHCb institutes. We acknowledge support from CERN and from the national agencies: CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and NSFC (China); CNRS/IN2P3 (France); BMBF, DFG and MPG (Germany); INFN (Italy); NWO (Netherlands); MEiN and NCN UMO-2018/31/B/ST2/03998 (Poland); MEN/IFA (Romania); MSHE (Russia); MICINN (Spain); SNSF and SER (Switzerland); NASU (Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the computing resources that are provided by CERN, IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), PIC (Spain), GridPP (United Kingdom), RRCKI and Yandex LLC (Russia), CSCS (Switzerland), IFIN-HH (Romania), CBPF (Brazil), PL-GRID (Poland) and NERSC (USA). We are indebted to the communities behind the multiple open-source software packages on which we depend. Individual groups or members have received support from ARC and ARDC (Australia); AvH Foundation (Germany); EPLANET, Marie Skłodowska-Curie Actions and ERC (European Union); A*MIDEX, ANR, IPhU and Labex P2IO, and Région Auvergne-Rhône-Alpes (France); Key Research Program of Frontier Sciences of CAS, CAS PIFI, CAS CCEPP, Fundamental Research Funds for the Central Universities, and Sci. & Tech. Program of Guangzhou (China); RFBR, RSF and Yandex LLC (Russia); GVA, XuntaGal and GENCAT (Spain); the Leverhulme Trust, the Royal Society and UKRI (United Kingdom). The authors would like to thank the LHCb computing and simulation teams for their support and for producing the simulated LHCb samples used to benchmark the performance of RTA software. Funding Information: We thank the technical and administrative staff at the LHCb institutes. We acknowledge support from CERN and from the national agencies: CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and NSFC (China); CNRS/IN2P3 (France); BMBF, DFG and MPG (Germany); INFN (Italy); NWO (Netherlands); MEiN and NCN UMO-2018/31/B/ST2/03998 (Poland); MEN/IFA (Romania); MSHE (Russia); MICINN (Spain); SNSF and SER (Switzerland); NASU (Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the computing resources that are provided by CERN, IN2P3 (France), KIT and DESY (Germany), INFN (Italy), SURF (Netherlands), PIC (Spain), GridPP (United Kingdom), RRCKI and Yandex LLC (Russia), CSCS (Switzerland), IFIN-HH (Romania), CBPF (Brazil), PL-GRID (Poland) and NERSC (USA). We are indebted to the communities behind the multiple open-source software packages on which we depend. Individual groups or members have received support from ARC and ARDC (Australia); AvH Foundation (Germany); EPLANET, Marie Sk?odowska-Curie Actions and ERC (European Union); A*MIDEX, ANR, IPhU and Labex P2IO, and R?gion Auvergne-Rh?ne-Alpes (France); Key Research Program of Frontier Sciences of CAS, CAS PIFI, CAS CCEPP, Fundamental Research Funds for the Central Universities, and Sci. & Tech. Program of Guangzhou (China); RFBR, RSF and Yandex LLC (Russia); GVA, XuntaGal and GENCAT (Spain); the Leverhulme Trust, the Royal Society and UKRI (United Kingdom). The authors would like to thank the LHCb computing and simulation teams for their support and for producing the simulated LHCb samples used to benchmark the performance of RTA software. Publisher Copyright: © 2021, The Author(s).

PY - 2022/12

Y1 - 2022/12

N2 - The Large Hadron Collider beauty (LHCb) experiment at CERN is undergoing an upgrade in preparation for the Run 3 data collection period at the Large Hadron Collider (LHC). As part of this upgrade, the trigger is moving to a full software implementation operating at the LHC bunch crossing rate. We present an evaluation of a CPU-based and a GPU-based implementation of the first stage of the high-level trigger. After a detailed comparison, both options are found to be viable. This document summarizes the performance and implementation details of these options, the outcome of which has led to the choice of the GPU-based implementation as the baseline.

AB - The Large Hadron Collider beauty (LHCb) experiment at CERN is undergoing an upgrade in preparation for the Run 3 data collection period at the Large Hadron Collider (LHC). As part of this upgrade, the trigger is moving to a full software implementation operating at the LHC bunch crossing rate. We present an evaluation of a CPU-based and a GPU-based implementation of the first stage of the high-level trigger. After a detailed comparison, both options are found to be viable. This document summarizes the performance and implementation details of these options, the outcome of which has led to the choice of the GPU-based implementation as the baseline.

KW - Heterogeneous

KW - High-level trigger

KW - High-throughput

KW - Parallel computing

KW - Real-time

KW - Software

UR - http://www.scopus.com/inward/record.url?scp=85122303493&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85122303493&partnerID=8YFLogxK

U2 - 10.1007/s41781-021-00070-2

DO - 10.1007/s41781-021-00070-2

M3 - Article

AN - SCOPUS:85122303493

SN - 2510-2044

VL - 6

SP - 1

EP - 20

JO - Computing and Software for Big Science

JF - Computing and Software for Big Science

IS - 1

M1 - 1

ER -

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

Abstract

Bibliographical note

Funding

Keywords

Access to Document

Persistent URL (handle)

Other files and links

Fingerprint

Cite this