基于GPU的加锁并行化非结构网格生成方法研究

Locking paralleled GPU-based method research for unstructured mesh generation

非结构网格的生成在时间和内存上有一定的缺陷,这里提出了一种新的方法,命名为GPU-PDMG,是基于CUDA架构的GPU并行非结构网格生成技术。该技术结合了GPU的高速并行计算能力与Delaunay三角化的优点,在英伟达GPU模块下采用CUDA程序模型,开发出了加锁并行区划分技术,通过对NACA0012翼型、多段翼型等算例进行测试,分析此方法的加速比和效率,对其计算性能展开评估。实验结果表明,GPU-PDMG优于现存在的CPU算法的速度,在保证网格质量的同时,提高了效率。

Defects of consuming time and memory consist in unstructured mesh generation. This paper proposes a novel approach, terming GPU-PDMG, which is GPU parallel unstructured mesh generation based on the framework of CUDA. The technology combines the high-speed parallel GPU and advantages of Delaunay triangulation. It develops a method of locking parallel area dividing, using the CUDA programming model on nVidia GPUs. By analyzing the tested examples’ speedup rate and efficiency, it has evaluated their computing performance. This result is identified in NACA0012 and multi-element airfoil experiment with both the analysis of speedup rate and efficiency and GPU-PDMG is better than any existing GPU algorithms.