Simple Geometry Compression for Ray Tracing on GPU
Keldysh Institute of Applied Mathematics, Russian Academy of Sciences
Proceedings of Graphicon, 2011
@article{garanzha2011simple,
title={Simple Geometry Compression for Ray Tracing on GPU},
author={Garanzha, K. and Bely, A. and Galaktionov, V.},
year={2011}
}
In this short paper we describe simple approach to loosely compress the vertex data for ray tracing large triangular models on the GPU. The disadvantage of the GPU is limited memory capacity. The advantage of the GPU is high performance computation. Sometimes it is hard to load large models to the GPU and we suggest compressing the vertices that form 3D models and acceleration data structure using 16bit representations instead of 32bit floats. At the same time the aggregate vertex precision varies between 22 and 24 bits and the amount of cracks is minimized. The advantages of NVIDIA CUDA platform are used to implement our approach efficiently.
December 26, 2011 by hgpu
