Compression methods for GPUs

(01.07.2009 - 30.06.2011)

Modern dedicated graphics processors (GPUs) make it possible to render complex scenes and objects in realtime. Currently, the bottle neck is often not rendering performance, but the available video memory. Goal of the project, which is done in collaboration with RTT, Munich, is to develop and examine methods for the reduction of memory consumption of scene data in video memory of a GPU.

Papers at LGDV

On Floating-Point Normal Vectors
On Floating-Point Normal Vectors
Memory Efficient Light Baking
Memory Efficient Light Baking
Multiresolution Attributes for Tessellated Meshes
Adaptive Level-of-Precision for GPU-Rendering
Adaptive Level-of-Precision for GPU-Rendering
Memory Efficient Light Baking

Promoter

  • RTT München