Portal de Periódicos da CAPES

3D Terrain Real-time Rendering Method Based on CUDA-OpenGL Interoperability

2015; Taylor & Francis; Volume: 32; Issue: 6 Linguagem: Inglês

10.1080/02564602.2015.1040473

0974-5971

Jiaji Wu, Long Deng, Anand Paul,

Geological Modeling and Analysis

In the field of geographic information, fast surface modelling techniques are important. Usually, mesh model is used to describe the terrain. However, the large data of mesh model are a bottleneck of real-time rendering. Compared with the CPU, GPU has more execution units and memory units, which greatly improves its computing capacity and memory bandwidth. Meanwhile, in the field of image rendering, OpenGL has always played an important role. Since CUDA (compute unified device architecture) and OpenGL both run on GPU and share data through common memory, this paper gives a three-dimensional terrain real-time rendering method based on CUDA–OpenGL interoperability. First, we use CUDA C kernel function to calculate the vertex coordinates and normal vectors, then we pass the data to OpenGL buffer, and render it. In this paper, we use share memory and register, overlap strategy, the expansion of concurrency and the reasonable block size. The experimental results show that this algorithm can improve the computing speed greatly, with the speedup to 212.3x, meeting the needs of real-time rendering.