"Graphical Optimization Heuristics for 3D Game Engines"

Linda Barber 2003


Abstract

The purpose of this IS is to investigate graphical optimization heuristics for 3D game engines. This is a complex and hierarchical topic, which includes quite a number of interesting theoretical issues. The first topic to be investigated is camera manipulation and the differences between first and third person camera views. Though the topic of camera manipulation has no direct impact on graphical optimization, it is difficult to discuss game engine design without it. The second subject is frustum culling and bounding volumes. Bounding volumes segue nicely into the third topic which is collision detection, a necessity in virtually every game engine. It is difficult to discuss large scale collision detection of many objects without examining spatial partitioning, so this issue is also discussed. Each of these topics is examined separately with OpenGL programs designed to test that particular concept.


cull1


Frustum culling is turned off and 330 of 330 tetrahedrons (with spherical bounding volumes) are processed in the rendering pipeline even though only 13 tetras could be visible.




cull2


When frustum culling is activated only those tetrahedrons within the view volume, 8 of 330, are processed in the rendering pipeline.