The connection between particle physics theory and experiment hinges on the simulation of detector events by means of Monte-Carlo methods. Event generators are central to the success of the accelerator-neutrino program, which will provide stringent tests of the three-flavor paradigm. A quantitatively correct modeling of neutrino interactions entails multiple energy scales and becomes computationally hard when including a fully quantum mechanical treatment of nuclear many-body effects, as may be needed to reach the precision targets at the Deep Underground Neutrino Experiment (DUNE). Event generators are also crucial components of the Large Hadron Collider (LHC) software stack. They can nowadays perform parton level simulations at high precision, which are subsequently matched to parton showers to provide particle-level events. With the upcoming data increase at the LHC, the computational footprint of precision event generators will grow dramatically, contributing to the outpaced growth in computing needs compared to available budgets. Event generation is a non-negligible fraction of LHC computing, and improved performance will help to reduce computational requirements.

This project aims to address the computing problem by constructing a performance portable solution that optimizes basic math operations needed in parton- and particle-level event gen- erators for a variety of architectures, parallelizes calculations across events, and capitalizes on heterogeneous systems. First applications will be a novel theory-driven event generator for the DUNE experiment, and the Monte Carlo for FeMtobarn processes (MCFM) for the LHC. This will serve as a blueprint for particle-level event generators and automated matrix-element generators. Guided by the known matching and merging algorithms in collider physics, the project will explore various possibilities to extend the current classical cascade models in neutrino event generators to a partial quantum cascade, for which computational efficiency is an essential prerequisite.

We build on capabilities in the FASTMath and RAPIDS SciDAC Institutes.

NeuCol is grateful to be supported by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research and Office of High-Energy Physics, Scientific Discovery through Advanced Computing (SciDAC) program under contracts DE-AC02-06CH11357 (Argonne), DE-AC02-07CH11359 (Fermilab), and DE-AC05-00OR22725 (Oak Ridge).

Code of Conduct

All collaborators are expected to be respectful of one another. Discussions and disagreements should remain polite and free of abusive language.

There will no discrimination based on gender, ethnicity, race, sexual orientation, or any other potentially divisive reason.

All major code components will have an associated list of authors.

Proper credit must be given to the contributors of code components used in all scientific publications.

All code contributors will follow the contribution policies and workflow and the coding standards.