Optimizing the management of continuous level of detail models on GPU

Abstract

In this paper we present a new continuous multiresolution framework which has been developed in view of the outstanding evolution of hardware. Our interest not only focuses on exploiting GPUs possibilities, but also on making the best possible use of the capabilities offered by new bus technologies. On the one hand, we store the geometry, based on triangle strips, in high-performance memory on the GPU, offering fast rendering time. On the other hand, we have designed our level-of-detail extraction algorithm in order to make the most of current PCI Express bus characteristics, by sending the minimum information in the most appropriate way while taking into account the appearance of degenerate triangles. The results section shows that our model improves the efficiency of previously existing solutions. Its easy integration and its short extraction time make it suitable for game engines and graphics libraries which often resort to discrete models when it comes to selecting a multiresolution technique. © 2008 Elsevier Ltd. All rights reserved.