基于相似度引导的流线种子点并行分布方法

Parallel Strategy of Streamline Seeding with Similarity-guiding

流线是矢量场可视化的重要方法之一,现有的基于相似度引导的流线分布方法能较好体现矢量场特征,但运行速度缓慢。为解决此问题,提出了一种基于相似度引导流线种子点分布的并行优化方法,该方法生成备选种子点集,各线程并行获取种子点,并积分生成流线。由于流线间相互影响,采用副本和缓存技术来避免线程间的读写冲突和等待问题,可得到满足相似距离约束的流线分布。实验结果表明,该方法能很好利用多核的并行计算优势,获得较高的并行加速比,有效提高流线的生成速度。

Streamline is one of the most important methods of vector field visualization. The streamline visualization describing and analyzing vector field feature is significant in scientific research. Existing similarity-guided streamline placement methods can well reflect vector field feature, but they are running at a low speed. A parallel strategy of streamline seeding with similarity-guiding was presented to solve the problem. First a seedpoint set was generated, and then seedpoints in the set were transferred to threads in parallel, and integrated to grow the streamline. In consideration of interaction between streamlines, copy and buffer technology was implemented to avoid read-write conflict and block-waiting problem. Finally the streamline placement that meets the constraint of similarity distance was produced. Experiment shows that the algorithm well takes advantage of multi-core parallel calculation, and reaches high speedup from paralleling, speeding up the generation of streamline efficiently.