PreprintCC BY 4.0arXiv
MadVfold: accelerating NLO event generation and reducing negative weights with SIMD vectorization and GPUs
Abstract
NLO simulations are essential for LHC physics analyses but are expensive, as they are not only slow but also lead to negative weights, which imply the need to simulate much larger samples of events. Folding is a powerful technique to reduce negative weights but is itself expensive. In this paper I propose ``vectorized folding'' as a new idea to speed up these calculations using SIMD and GPUs, and I present its CUDACPP-based implementation for MG5aMC in MadVfold, including its extension for unfolded NLO event generation. Preliminary results show overall speedups around 6x to 9x with folding and 3x without it. This work is based on a test-centric, LLM-assisted software development process.
Related
- The Bondi dipole in full numerical relativity: a self-accelerating positive-negative mass binary
- Accelerating noisy variational quantum algorithms with physics-informed denoising networks
- Fields on accelerating and super accelerating black holes: stability, quasinormal modes, greybody factors and strong cosmic censorship
- Is there an accelerating nonspreading wave packet in the Schrödinger equation with higher even-order dispersions?
- Theoretical necessity of the Accelerating Universe
BibTeX
@misc{valassi2026madvfold,
title = {{MadVfold: accelerating NLO event generation and reducing negative weights with SIMD vectorization and GPUs}},
author = {Andrea Valassi},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.23500},
eprint = {2609.23500},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.23500},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - MadVfold: accelerating NLO event generation and reducing negative weights with SIMD vectorization and GPUs
AU - Andrea Valassi
PY - 2026
DA - 2026/09/20
PB - arXiv
DO - 10.48550/arxiv.2609.23500
UR - https://arxiv.org/abs/2609.23500
AB - NLO simulations are essential for LHC physics analyses but are expensive, as they are not only slow but also lead to negative weights, which imply the need to simulate much larger samples of events. Folding is a powerful technique to reduce negative weights but is itself expensive. In this paper I propose ``vectorized folding'' as a new idea to speed up these calculations using SIMD and GPUs, and I present its CUDACPP-based implementation for MG5aMC in MadVfold, including its extension for unfolded NLO event generation. Preliminary results show overall speedups around 6x to 9x with folding and 3x without it. This work is based on a test-centric, LLM-assisted software development process.
ER -