MC event generators experiments integration

Tuesday 27 Oct 2026, 16:00 → 17:00 Europe/Zurich

513/R-070 - Openlab Space (CERN)

Meeting minutes 2026-10-27

Quick recap

The meeting focused on preparing materials for the upcoming WLCG/HSF workshop and reviewing the status of generator integration and validation for ATLAS and CMS.
Saptaparna reported that the physics content of Dipesh's plots had been approved by ATLAS, though some aesthetic changes and official labeling were still needed.
Daniele planned to use these plots in a plenary talk to show that experiments are starting to adopt GPU-based generators.
Lorenzo noted that CMS had no new public material for MadGraph7 but suggested using older public plots from an internal note and proceedings.
The group discussed the possibility of a common validation framework for generators across experiments, noting that different PDF choices and baseline samples between ATLAS and CMS make direct comparisons difficult.
Saptaparna mentioned that ATLAS had provided funding for Dipesh to work with Zach at Berkeley.
Enrico provided an update on PEPPER integration in Athena and shared preliminary performance results for Sherpa, showing potential speedups from offloading leading-order processes to GPUs.

Next steps

Dipesh

• Update the plots with the changes requested by Zach and Saptaparna, including adding "ATLAS simulation preliminary" labels and ensuring the generator name is consistent (e.g., "MadGraph5_aMC@NLO").
• Add a plot showing strong scaling (number of cores on the x-axis and time on the y-axis) for the CPU scenario.
• Rerun the DY+2j inclusive event generation on the NGT cluster to get a more stable curve and smaller error bars, ensuring all runs (including Fortran) are on the same hardware.
• Clarify in the plot captions that the event generation is performed using gridpacks.
• Fix aesthetic issues on the plots, such as adding commas every three zeros in large numbers.
• Send the updated plots to Saptaparna, Daniele, and Stefan in an email thread for final approval before the workshop.

Saptaparna

• Ask ATLAS if a statement can be made that they are beginning validation of MadGraph7.
• Continue pinging the relevant people and tracking the Jira ticket to get the PEPPER merge request included in a usable release.
• Update the integration slides with the latest merge request information showing that the full workflow (matrix element generation and Pythia) can now happen within Athena.

Daniele

• Use the approved MadGraph5+CUDACPP plots and the CMS public material (Jin's proceedings) in the WLCG/HSF Workshop plenary presentation.
• Mention in the plenary that ATLAS has provided funding for Dipesh to work with Zach at Berkeley on MadGraph.

Summary

WLCG/HSF Workshop Planning Updates

Saptaparna reported that the plots have been approved except for some requested changes like the ATLAS logo.
The team discussed plans for next week's WLCG/HSF Workshop in Bologna where a plenary presentation would showcase sites ready to run generators on GPU, but Saptaparna expressed caution about sharing the plots until they receive full approval from ATLAS.

Content Approval and Validation Updates

The team discussed progress on content approval and plot updates, with Saptaparna confirming that the content is approved and Dipesh should share updated plots with stakeholders as soon as possible.
Saptaparna expressed caution about making public statements regarding MadGraph7 validation until proper results are confirmed, though she indicated workshop experiments are aware and supportive of the validation process.
The team also discussed ongoing issues with release merging and testing for PEPPER and CMS software, with Saptaparna committing to follow up on the delayed release creation.

CMS Presentation Material Discussion

The team discussed using existing CMS public material for an upcoming presentation, particularly regarding MapGraph for GPU version (CUDACPP) work published in CHEP 2024 proceedings.
Lorenzo confirmed they could use this material and offered to share the links, while Saptaparna advised against showing CMS and ATLAS plots on the same slide to avoid potential questions.
The discussion concluded with an inquiry about progress on MadGraph7, though no updates were provided during the meeting.

MadGraph Testing and Collaboration Updates

Lorenzo reported that testing of MadGraph7 has begun but is still in early stages as new team members are learning the system.
He proposed involving a PhD student from the University of Florence in a MadGraph project at CERN, with potential collaboration between CMS and the other experiment.
Saptaparna mentioned that ATLAS Center funding has been secured for Dipesh to work with Zach at Berkeley on MadGraph-related work.
Ali noted that he presented to CMS the previous day and will soon begin performance tests again.

Monte Carlo Generator Validation Framework

The group discussed developing a common framework for validating Monte Carlo generators across CMS and ATLAS experiments, with Lorenzo proposing a platform where validation results and plots from different experiments could be shared and cross-checked.
Saptaparna emphasized the challenges in creating common samples due to different experimental approaches and PDF choices between CMS and ATLAS.
The team agreed that a common framework would benefit development efforts and provide a mechanism to see results from both experiments without relying on collaboration politics, though they acknowledged this would be a significant project requiring further discussion.

Plot Approvals and Presentation Updates

The team discussed plot approvals and updates for upcoming presentations.
Saptaparna confirmed that the physics content of Dipesh's plots is approved, but additional ATLAS labels and clarifications are needed.
Dipesh agreed to submit new jobs for better curve fitting, add ATLAS labels, specify inclusive jet definitions, and clarify that event generation uses GridPack.
The team also discussed Enrico's upcoming presentation at the Thursday ATLAS Generators meeting about PEPPER integration and speedup results, with a planned release of version 3.1 by end of November.

16:00 → 16:50 Discussion

Speaker: Dr. All