Parallel terrain rendering using a cluster of computers

Abstract

This article presents a distributed parallel processing technique for rendering massive terrain using a cluster of machines consisting of one designated rendering node and 20 computing nodes. With a novel approach, the presented technique achieves an increase in rendering speed and an improvement in rendering capability. Adaptive terrain mesh constructions are done in parallel at computing nodes and the resulting meshes are combined and subsequently rendered at the rendering node. This study uses a height field of the United States at 30-m resolution spacing. It is divided into smaller blocks consisting of 4096 × 4096 vertices. Each computing node is assigned one or four blocks and tasked with creating the level-of-detail mesh that corresponds to view-dependent parameters provided by the rendering node. These individual terrain meshes are subsequently combined and rendered as seamless terrain meshes with a continuous terrain surface. The high rendering capacity of the presented technique is essential to the high-resolution large display system.

Keywords:

• terrain visualization
• parallel rendering
• level-of-detail

Acknowledgments

This study was supported in part by the National Science Council of Taiwan, ROC, through grants NSC 99-2218-E-027-008, NSC 100-2218-E-027-002, NSC 100-2221-E-027-090, NSC 101-2218-E-027-001-, and NSC 101-2221-E-027-131-.