EPJ Web of Conferences (Jan 2024)

Madgraph5_aMC@NLO on GPUs and vector CPUs Experience with the first alpha release

  • Hageböck Stephan,
  • Childers Taylor,
  • Hopkins Walter,
  • Mattelaer Olivier,
  • Nichols Nathan,
  • Roiser Stefan,
  • Teig Jørgen,
  • Valassi Andrea,
  • Vuosalo Carl,
  • Wettersten Zenny

DOI
https://doi.org/10.1051/epjconf/202429511013
Journal volume & issue
Vol. 295
p. 11013

Abstract

Read online

Madgraph5_aMC@NLO is one of the most-frequently used MonteCarlo event generators at the LHC, and an important consumer of compute resources. The software has been reengineered to maintain the overall lookand-feel of the user interface while speeding up event generation on CPUs and GPUs. The most computationally intensive part, the calculation of “matrix elements”, is offloaded to new implementations optimised for GPUs and for CPU vector instructions, using event-level data parallelism. We present the work to support accelerated leading-order QCD processes, and discuss how this work is going to be released to Madgraph5_aMC@NLO’s users.