PDF

Keywords

GPU
CUDA
Edge-function
Parallel implementation
voxelization

Abstract

The voxelization process is an essential stage in three dimensional (3D) graphics pipeline. Its implementation should precede displaying objects in the pipeline. In this paper, different Voxelization algorithms are modified and parallelized to accelerate the operation of this stage. The 3D Digital Differential Analyzer (DDA) algorithm is used for line voxelization. This algorithm is utilized in triangle filling using the scan-line and the edge-function algorithms. The first one is designed to produce lines in parallel while the second can produce voxels. All these algorithms are parallelized using CUDA architecture and implemented on GPU processor. The actual implementation of these algorithms is examined and optimized according to the occupancy and block size metrics. The experimental results show that the acceleration amount of 3D DDA was about 4352x max compared to the OpenGL implementation, and the edge function implementation has been executed at a higher speed than the scan-line for object triangles voxelization.
https://doi.org/10.33899/rengj.2020.126643.1015
  PDF