2507.00622
Chaoticus: a parallel approach to the computation of chaos indicators
Javier Jiménez-López, José Sáez-Landete, V. J. García-Garrido
correctmedium confidence
- Category
- Not specified
- Journal tier
- Specialist/Solid
- Processed
- Sep 28, 2025, 12:56 AM
- arXiv Links
- Abstract ↗PDF ↗
Audit review
The paper empirically demonstrates GPU-side monotone scaling in time-per-IC with batch size up to device saturation, robust energy conservation (ΔH ≈ 10^−9 for the double pendulum and ≈10^−8 for the other tests), and linear GPU scaling in N for FPU+GALI4; the model provides a complementary, more formal throughput/error analysis that explains these observations. The paper lacks formal proofs and some implementation assumptions (e.g., scheduler/work-conservation) are implicit, but its claims align with the model’s derivations, which are correct under standard smoothness and scheduling hypotheses.
Referee report (LaTeX)
\textbf{Recommendation:} minor revisions
\textbf{Journal Tier:} specialist/solid
\textbf{Justification:}
The manuscript presents a useful GPU-accelerated package that integrates high-order ODE solvers with canonical chaos indicators and demonstrates strong empirical performance across three benchmarks. The observed scaling to GPU saturation, linear-in-N GPU complexity for FPU+GALI4, and robust energy conservation are all credible and align with standard throughput and error analyses. To strengthen reproducibility and generality, the paper should add explicit details on GPU occupancy/capacity, the adaptive controller settings, and quantitative CPU–GPU speedups.