基于CUDA的等值线云图并行生成算法

Parallel Generation Algorithm of Contour Nephogram Based on CUDA

为了提高不规则网格中等值线云图的计算速度,提出了一种基于CUDA的等值线云图并行生成算法.首先给出了不规则网格中等值线的串行生成算法,该算法循环处理每个网格单元与每条等值线的相交,不需要考虑同一条等值线与不同网格的交点的先后关系.其次给出了等值线云图的串行算法,循环处理每个网格单元,根据它被不同等值线分割的情况求得很多等值多边形,最终得到的相邻的等值多边形会自然拼出云图,而不需要考虑不同等值多边形的先后关系.然后给出了基于CUDA的等值线和等值线云图的并行算法,对算法中内存设计和线程设计进行了充分讨论,尽量提高加速比.最后利用多组数据进行验证,结果证明并行算法功能和串行算法完全一致,而加速比达到了10倍以上.

To improve the computation speed of contour nephograms in irregular mesh,a parallel contour nephogram generation algorithm based on CUDA was proposed.Firstly,a serial contour generation algorithm in irregular mesh was presented,which deals with the intersections of each mesh unit and each contour in a cyclic manner,without considering the sequential relationship between the intersections of the same contour and different mesh units.Secondly,the serial algorithm of contour nephograms was given,which circulates each grid element,and many equivalent polygons are obtained according to the situation that it is divided by different contours.The final result adjacent equivalent polygons will naturally form the cloud map without considering the sequence relationship of different equivalent polygons.Then,the parallel algorithms of contour and contour nephogram based on CUDA were given,and the memory design and thread design in the algorithms are fully discussed to improve the acceleration ratio as far as possible.Finally,several groups of data are used to verify the results.The results show that the function of the parallel algorithm is exactly the same as that of the serial algorithm,and the acceleration ratio reaches more than 10 times.