
18. Hierarchical Depth Culling and Bounding-Box Management on the GPU 257
18.5 Conclusion and Future Work
I have presented a culling pipeline that effectively marks occluded objects according
to the depth buffer mipmap set. The pipeline runs mostly on the GPU and requires
minimal DMA transfer. It is designed to support skinning, morphing, and any other
GPU-side mesh modifications. This feature is helpful for graphics engines that try
not to keep synchronized copies of meshes in main memory and perform all mesh-
related transformations on the GPU.
Using hierarchical depth culling makes classical frustum culling partially obso-
lete. Objects outside the ...