摘要
提出一种完全基于GPU(graphics processing unit)的高度并行Marching Cubes改进算法.针对Marching Cubes算法流程中数据处理阶段进行优化.首先并行遍历每个体元,以序列形式得到每体元的非空状态,再利用高效流式缩减操作获得非空体元序列以便仅针对非空体元高度并行地生成等值面三角形.实验表明,与同市场价格CPU实现相比,该算法效率最高提高到了前者的9倍以上.
We present a fully GPU-based highly parallel Marching Cubes algorithm.The method mainly focuses on the data-processing stage to optimize the algorithm.Firstly,GPU use its highly parallel characters to traverse all voxels,and form non-empty state sequence.Secondly,we realize high-efficient stream reduction to achieve the sequence of the non-empty elements for triangles generation.The results show the speed of our algorithm can be improved by up to more than 9 times compared to same price CPU-based implementations.
出处
《微电子学与计算机》
CSCD
北大核心
2008年第9期151-154,157,共5页
Microelectronics & Computer