AG-2025.03-1717·physics.comp-ph·cross-listed: hep-exhep-ph
Madgraph on GPUs and vector CPUs: towards production (The 5-year journey to the first LO release CUDACPP v1.00.00)
Authors
- Andrea Valassi
- Taylor Childers
- Stephan Hageböck
- Daniele Massaro
- Olivier Mattelaer
- Nathan Nichols
- Filip Optolowicz
- Stefan Roiser
- Jørgen Teig
- Zenny Wettersten
Abstract
The effort to speed up the Madgraph5_aMC@NLO generator by exploiting CPU vectorization and GPUs, which started at the beginning of 2020, has delivered the first production release of the code for leading-order (LO) processes in October 2024. To achieve this goal, many new features, tests and fixes have been implemented in recent months. This process benefitted also from the early feedback of the CMS experiment. In this contribution, we report on these activities and on the status of the LO software at the time of CHEP2024.
Submitted
27 March 20251 year ago
Version
v1
License
CC-BY-4.0
DOI
10.48550/arXiv.2503.21935
Summary
Physicists have completed a five-year effort to make Madgraph5_aMC@NLO—a crucial tool for simulating particle collisions—run fast on GPUs and modern CPUs, releasing the first production-ready version in October 2024.
- Madgraph is essential software that calculates what happens when particles collide at the Large Hadron Collider; accelerating it on GPUs and vector CPUs makes these calculations orders of magnitude faster.
- The team incorporated real feedback from CMS, one of the LHC's major experiments, ensuring the code works for actual physics analysis rather than just in theory.
- This release enables particle physicists to run more simulations and explore more collision scenarios within practical compute budgets—crucial for discovering rare events or measuring subtle effects.
curious · generated by claude-haiku-4-5
Chat with this PDF
Ask questions, probe assumptions, request a plain-English summary. Answers cite sections from the preprint itself.
Community
Questions and answers about this paper from other readers. No formal peer review — just a place to think out loud.