Challenges in Monte Carlo Event Generator Software for High-Luminosity LHC

Valassi, A.; Yazgan, E.; Mcfayden, J. A.; Amoroso, S.; Bendavid, J.; Buckley, A. G.; Cacciari, Matteo; Childers, J. T.; Ciulli, V.; Frederix, Rikkert; Frixione, Stefano; Giuli, F.; Grohsjean, A.; Gütschow, C.; Höche, Stefan; Hopkins, W.; Ilten, P.; Konstantinov, D.; Krauss, Frank; Li, Qiang; Lönnblad, Leif; Maltoni, Fabio; Mangano, M.; Marshall, Z.; Mattelaer, Olivier; Menéndez, J. Fernández; Mrenna, S.; Muralidharan, Servesh; Neumann, Tobias; Plätzer, Simon; Prestel, Stefan; Roiser, S.; Schönherr, Marek; Schulz, H.; Schulz, M.; Sexton-Kennedy, E.; Siegert, F.; Siódmok, Andrzej; Stewart, G. A.

doi:10.1007/s41781-021-00055-1

Cited by 40 publications

(37 citation statements)

References 152 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The very first step in our reengineering of MG5aMC, and in the design of a software architecture enabling efficient parallel processing on both GPUs and CPUs, was to deconstruct its computational workflow into separate components and to identify their input and output data. This phase of the project, which overlapped with the work ongoing in the HSF Event Generator WG for preparing the LHCC review of HL-LHC computing [2,3], was also key to enabling the collaboration between people with diverse skills and backgrounds, by translating theoretical physics methodologies into practical computational problems.…”

Section: Component Decomposition Of Matrix Element Generators -Event-level Data Parallelismmentioning

confidence: 99%

“…As described more in detail in Refs. [2,3], a Matrix Element Generator (MEG) like MG5aMC is a computer program that, starting from a limited set of input configuration parameters, is typically used to produce two types of outputs: a fiducial cross section for the collision process and a sample of N unw unweighted events. Both results are obtained using Monte Carlo methods, starting from a larger set of N wgt randomly generated weighted events.…”

Section: Component Decomposition Of Matrix Element Generators -Event-level Data Parallelismmentioning

confidence: 99%

“…MadGraph5_aMC@NLO [1] (in the following, MG5aMC) is a physics event generator software used in the data processing workflows of High Energy Physics (HEP) experiments, such as ATLAS and CMS at CERN's Large Hadron Collider (LHC). While the majority of CPU resources in the distributed computing environments of the LHC experiments is spent on detector simulation and event reconstruction workflows, physics event generators are also large consumers of CPU time, accounting for an estimated 12% of overall CPU budgets for ATLAS and 5% for CMS [2]. Their computational cost is also predicted to increase during the High-Luminosity phase of LHC (HL-LHC), as more accurate theoretical predictions are required for physics analyses.…”

Section: Introductionmentioning

confidence: 99%

“…Their computational cost is also predicted to increase during the High-Luminosity phase of LHC (HL-LHC), as more accurate theoretical predictions are required for physics analyses. Reengineering event generators to improve their software performance in view of the HL-LHC, especially on modern hardware architectures such as CPUs with many cores and wide vector registers, is therefore essential [2,3].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Design and engineering of a simplified workflow execution for the MG5aMC event generator on GPUs and vector CPUs

et al. 2021

Self Cite

View full text Add to dashboard Cite

Physics event generators are essential components of the data analysis software chain of high energy physics experiments, and important consumers of their CPU resources. Improving the software performance of these packages on modern hardware architectures, such as those deployed at HPC centers, is essential in view of the upcoming HL-LHC physics programme. In this paper, we describe an ongoing activity to reengineer the Madgraph5_aMC@NLO physics event generator, primarily to port it and allow its efficient execution on GPUs, but also to modernize it and optimize its performance on vector CPUs. We describe the motivation, engineering process and software architecture design of our developments, as well as the current challenges and future directions for this project. This paper is based on our submission to vCHEP2021 in March 2021, complemented with a few preliminary results that we presented during the conference. Further details and updated results will be given in later publications.

show abstract

Section: Component Decomposition Of Matrix Element Generators -Event-level Data Parallelismmentioning

confidence: 99%

Section: Component Decomposition Of Matrix Element Generators -Event-level Data Parallelismmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Design and engineering of a simplified workflow execution for the MG5aMC event generator on GPUs and vector CPUs

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…While the best methods exhibit good scaling with the number of final-state particles and are the only means to perform very high multiplicity calculations, it is prudent to resort to known analytic results whenever they are available and computational resources are scarce. The problem has become pressing due to the fact that the computing power on the Worldwide LHC Computing Grid (WLCG) is projected to fall short of the demand by at least a factor two in the high-luminosity phase of the Large Hadron Collider (LHC) [22][23][24][25]. Moreover, most techniques for fully differential NNLO calculations rely on the fast and numerically stable evaluation of one-loop results in infrared-singular regions of phase space, further increasing the demand for efficient one-loop computations [26,27].…”

Section: Introductionmentioning

confidence: 99%

Accelerating LHC phenomenology with analytic one-loop amplitudes

2021

Self Cite

View full text Add to dashboard Cite

The evaluation of one-loop matrix elements is one of the main bottlenecks in precision calculations for the high-luminosity phase of the Large Hadron Collider. To alleviate this problem, a new C++ interface to the "Image missing" parton-level Monte Carlo is introduced, giving access to an extensive library of analytic results for one-loop amplitudes. Timing comparisons are presented for a large set of Standard Model processes. These are relevant for high-statistics event simulation in the context of experimental analyses and precision fixed-order computations.

show abstract

Accelerating HEP simulations with Neural Importance Sampling

Deutschmann,

Götz

2024

J. High Energ. Phys.

View full text Add to dashboard Cite

Many high-energy-physics (HEP) simulations for the LHC rely on Monte Carlo using importance sampling by means of the VEGAS algorithm. However, complex high-precision calculations have become a challenge for the standard toolbox, as this approach suffers from poor performance in complex cases. As a result, there has been keen interest in HEP for modern machine learning to power adaptive sampling. While previous studies have shown the potential of normalizing-flow-powered neural importance sampling (NIS) over VEGAS, there remains a gap in accessible tools tailored for non-experts. In response, we introduce ZüNIS, a fully automated NIS library designed to bridge this divide, while at the same time providing the infrastructure to customise the algorithm for dealing with challenging tasks. After a general introduction on NIS, we first show how to extend the original formulation of NIS to reuse samples over multiple gradient steps while guaranteeing a stable training, yielding a significant improvement for slow functions. Next, we introduce the structure of the library, which can be used by non-experts with minimal effort and is extensivly documented, which is crucial to become a mature tool for the wider HEP public. We present systematic benchmark results on both toy and physics examples, and stress the benefit of providing different survey strategies, which allows higher performance in challenging cases. We show that ZüNIS shows high performance on a range of problems with limited fine-tuning.

show abstract

Challenges in Monte Carlo Event Generator Software for High-Luminosity LHC

Cited by 40 publications

References 152 publications

Design and engineering of a simplified workflow execution for the MG5aMC event generator on GPUs and vector CPUs

Design and engineering of a simplified workflow execution for the MG5aMC event generator on GPUs and vector CPUs

Accelerating LHC phenomenology with analytic one-loop amplitudes

Accelerating HEP simulations with Neural Importance Sampling

Contact Info

Product

Resources

About