Omnisolver: An extensible interface to Ising spin-glass and QUBO solvers: adding a distributed GPU brute-force plugin
Abstract
This software update extends Omnisolver with \texttt{omnisolver-bruteforce}, a first-class plugin for exact exhaustive search of QUBO and Ising instances on CUDA-enabled GPUs. The update contributes three components: packaging of the single-GPU brute-force kernel of [Computer Physics Communications 260, 107728, 2021] as a first-class Omnisolver plugin with a uniform Python and CLI interface, a Ray-based distributed framework that splits the search into 2k fixed-variable subproblems and dispatches them across multiple GPUs and hosts, with a controller that gathers and merges partial results, and numerical stabilization of the fast \texttt{float32} ground-state path through compensated incremental updates, periodic exact energy re-anchoring, and best-buffer refresh. Stabilization activates automatically for N ≥ 40 leaving the public sampler API is unchanged. On dense random Ising instances we measure exhaustive solve times up to N=60 on 8× NVIDIA H100 (96 GB) GPUs, reaching ≈ 3.15 days at N=60 in close agreement with the empirical tN+1=2 tN doubling rule, with the distributed sampler reaching the full ≈ 8× speedup over a single H100 from N ≳ 44 onwards. The plugin thereby serves as a practical ground-truth oracle for state-of-the-art classical heuristics such as the Simulated Bifurcation Machine.
Related
- ESPResSo++: A Fast and Extensible Molecular Simulation Package for Coarse-Grained Models
- QuantumBilliards.jl: An extensible, user friendly, high-performance Julia library for the large scale computation of spectra and eigenstates of quantum billiards and related problems
- Extensible links in a broad class of single polymer chain models
- An Extensible Agentic Framework for Expert-Level Automation of Atomistic Simulations
- GWPTools: An extensible Wolfram Language framework for time-dependent Gaussian wave packet dynamics
BibTeX
@misc{jalowiecki2026omnisolver,
title = {{Omnisolver: An extensible interface to Ising spin-glass and QUBO solvers: adding a distributed GPU brute-force plugin}},
author = {Konrad Jałowiecki and Jakub Pawłowski and Bartłomiej Gardas and Łukasz Pawela},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.10542},
eprint = {2610.10542},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.10542},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Omnisolver: An extensible interface to Ising spin-glass and QUBO solvers: adding a distributed GPU brute-force plugin
AU - Konrad Jałowiecki
AU - Jakub Pawłowski
AU - Bartłomiej Gardas
AU - Łukasz Pawela
PY - 2026
DA - 2026/10/01
PB - arXiv
DO - 10.48550/arxiv.2610.10542
UR - https://arxiv.org/abs/2610.10542
AB - This software update extends Omnisolver with \texttt{omnisolver-bruteforce}, a first-class plugin for exact exhaustive search of QUBO and Ising instances on CUDA-enabled GPUs. The update contributes three components: packaging of the single-GPU brute-force kernel of [Computer Physics Communications 260, 107728, 2021] as a first-class Omnisolver plugin with a uniform Python and CLI interface, a Ray-based distributed framework that splits the search into 2^k fixed-variable subproblems and dispatches them across multiple GPUs and hosts, with a controller that gathers and merges partial results, and numerical stabilization of the fast \texttt{float32} ground-state path through compensated incremental updates, periodic exact energy re-anchoring, and best-buffer refresh. Stabilization activates automatically for N ≥ 40 leaving the public sampler API is unchanged. On dense random Ising instances we measure exhaustive solve times up to N=60 on 8× NVIDIA H100 (96 GB) GPUs, reaching ≈ 3.15 days at N=60 in close agreement with the empirical t_N+1=2 t_N doubling rule, with the distributed sampler reaching the full ≈ 8× speedup over a single H100 from N ≳ 44 onwards. The plugin thereby serves as a practical ground-truth oracle for state-of-the-art classical heuristics such as the Simulated Bifurcation Machine.
ER -