SDOF-Based Algorithm for Parallel Generation of P-I Curves
Jie Liu, Li‐Dan Kuang, Shu-Meng Lei · 2025
The P-I curve is a crucial tool for assessing structural blast damage and is widely applied in the fields of disaster mitigation, prevention, and weapon effects evaluation. Traditional methods for generating the P-I curve typically require multiple equivalent single-degree-of-freedom (SDOF) calculations. However, serial algorithms struggle to meet the demand for rapid P-I curve generation across multiple building components. To enhance the efficiency of $\mathbf{P}-\mathbf{I}$ curve generation, this paper presents an optimization of the P-I curve algorithm using -O3 compilation flags, followed by further optimization through OpenMP multithreading parallelization. Performance experiments of the parallel P-I curve generation algorithm were conducted on the Tianhe next-generation system, focusing on the parallel performance of the algorithm across varying data sizes. The experimental results indicate that the compilationoptimized serial P-I curve generation algorithm achieves a speedup ratio of 2.2 compared to the original algorithm, with performance improvements being independent of data size. The parallel performance of the P-I curve generation algorithm shows a positive correlation with data size, with 32-thread parallelism demonstrating the highest efficiency. For a dataset of size $10,000 \times 10,000$, the 32-thread parallel efficiency reaches $107.36 \%$ compared to single-thread performance, achieving superlinear parallel performance. These findings suggest that the optimized algorithm can effectively support the parallel generation of P-I curves for building components in real-world applications.