基于CUDA的Dividing Cubes算法实现
Implementation of Dividing Cubes Algorithm Based on CUDA
-
摘要: 针对传统Dividing Cubes算法中一次性将体元剖分为较小体元而带来大量计算的问题，提出递进式的体元剖分方式，以减少剖分过程中的计算量。同时，将算法移植到图形处理器（GPU）上执行，在统一计算设备架构（CUDA）下对其进行实现，并分别对 kernel函数、线程结构和存储空间进行了设计和分配。实验结果表明，利用GPU的并行加速能力可以达到将近10倍的加速比，有效地提升了等值面的绘制速度。Abstract: In traditional Dividing Cubes algorithm,voxels are divided into many small elements in one time which bring a large number of calculations.Therefore,a progressive subdivision is proposed to reduce calculations.In the meanwhile,the algorithm is transplanted into GPU, and implemented based on CUDA. The kernel, thread organization and memory allocation is also designed.Experimental results show that almost 10 times speedup is reached by using GPU's parallel acceleration capacity.
下载:
