Efficient Ray Traversal of Constrained Delaunay Tetrahedralization
Efficient Ray Traversal of Constrained Delaunay Tetrahedralization
复制标题
约束 Delaunay 四面体化的高效光线遍历
DOI：
--
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
L. Aveneau
中科院分区：
文献类型：
--
作者：
M. Maria;S. Horna;L. Aveneau
加速结构对于光线跟踪应用程序是强制性的，允许每秒投射大量光线。2008年，Lagae和Dutre提出使用约束Delaunay四面体（CDT）作为光线跟踪的加速结构。我们的实验表明，他们的遍历算法是不适合GPU的应用程序，主要是由于算法错误。本文提出了一种新的CDT遍历算法。该算法比以前的算法效率更高：它使用更少的算术运算;由于它使用了固定数量的运算，所以它没有增加额外的线程发散;最后，与以前的遍历算法相反，它对32位浮点数具有鲁棒性。因此，它是第一个在CPU和GPU上都可用的方法。
Acceleration structures are mandatory for ray-tracing applications, allowing to cast a large number of rays per second. In 2008, Lagae and Dutre have proposed to use Constrained Delaunay Tetrahedralization (CDT) as an acceleration structure for ray tracing. Our experiments show that their traversal algorithm is not suitable for GPU applications, mainly due to arithmetic errors. This article proposes a new CDT traversal algorithm. This new algorithm is more efficient than the previous ones: it uses less arithmetic operations; it does not add extra thread divergence since it uses a fixed number of operation; at last, it is robust with 32-bits floats, contrary to the previous traversal algorithms. Hence, it is the first method usable both on CPU and GPU.