Virtual texturing project pitch

A -very- brief outline of my proposed virtual texturing project.

Transcript

1. Stuart Milne Bsc (Hons) Computer Games Technology Virtual texturing in real time applications
2. The Topic Virtual texturing “Virtual memory for textures” Increase texture detail Dynamic texture streaming Decreased memory cost Dynamic level of detail Used in Rage and Brink Rage had 128k x 128k textures! Environment/terrain focus
3. The Question “How can virtual texturing be used to increase visual fidelity in a real time application?” Aim To implement virtual texturing technology. To assess the visual fidelity benefits afforded by a virtual texturing system in a real time application.. Objectives To research current virtual texturing implementations Implement a virtual texturing system for use in a real-time application To explore possible optimisations utilising GP-GPU methods Demonstrate the virtual texturing system by utilising it to produce a real-time scene Compare the virtually textured scene against a comparable traditional implementation Measure the performance characteristics of the virtual texturing system Compare the performance characteristics against a traditional approach
4. The Literature Tanner, C., et al. 1998. The Clipmap : A Virtual Mipmap The beginnings – An overview and implementation of clipmapping by SGI. This is the beginning of virtual texturing. Ed. Engel, W. GPU Pro. A K Peters. ‘Virtual Texturing 101’and ‘Accelerating Virtual Texturing using CUDA’ chapters; the clues in the name! Great overview of problem space. GPU acceleration is a possible ‘focus’. Waveren, J.M.P. 2009. Id Tech 5 Challenges – From texture virtualisation to massive parallelisation Senior programmer id software @ Siggraph ‘09. Some specifics from RAGE. Filtering, cache thrashing, LOD ‘popping’ covered briefly. ‘Real-world’ insight. Further reading: Virtual Memory -> Sparse Virtual Textures -> Parallelisation
5. The Methodology Practical work Big show-off scene LOD is key - plenty of scale Technology visualisation – ability to see under the hood Evaluation Subjective comparisons – v.tex vs. traditional textures Scale should speak for itself! Quantative data – profile and statistics Memory usage/savings Internal profiling – paging efficiency, cache effectiveness etc..