1 / 18

Realishtime Radiosity And Next Shooter Selection

Realishtime Radiosity And Next Shooter Selection. Jeff Pool COMP 870 Final December 4, 2008 UNC Dept. of Computer Science. Approaches to Radiosity. CPU matrix solving GPU gathering GPU scattering GPU scatter/gathering Hemicubes Hemispheres. I tried:. GPU gathering Hemicubes

remedy
Download Presentation

Realishtime Radiosity And Next Shooter Selection

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. RealishtimeRadiosityAnd Next Shooter Selection Jeff Pool COMP 870 Final December 4, 2008 UNC Dept. of Computer Science

  2. Approaches to Radiosity • CPU matrix solving • GPU gathering • GPU scattering • GPU scatter/gathering • Hemicubes • Hemispheres

  3. I tried: • GPU gathering • Hemicubes • 5 renders per subelement • No form factor calculations, though • FBOs • Hemispheres • Scatter/Gather

  4. Approach 1 - Hemicubes • Gathering • 50x50 elements • Hemicubes • 5 renders per element • CPU processing • Bandwidth issues  ~24 hours

  5. Approach 2 – Hemicubes w/ FBO • Gathering • Hemicubes • FBO • No sending data back and forth ~5 minutes

  6. Approach 3 – Coombe, et al. • Scatter/Gather • Hemispheres • FBO  ~1/2 second

  7. Key Points • ID Buffer • Hemispheres • Progressive refinement • Next shooter selection • (fpMipMap generation) • Texture packing for residual/full textures • (HW now has MRT)

  8. ID Buffer • Elements rendered with ID colors for visibility (hard part of form factors) • Correlated to hemisphere rendered by shooter

  9. Hemispheres • Only one pass, not 5 • Can be some distortion, so target framebuffer must be large for accurate visibility

  10. ID texture size

  11. Shooting Pass • Scene rendered into hemisphere by shooter • Each element is rendered as a quad • Per fragment: • Form factor computation • Visibility = Idcolor==hemisphere(projectedPos) • Energy = FF * Reflectance * Visibility

  12. Updating the elements • Easy: Blend 1 + 1 (src + dst) = additive rendering • Render to multiple targets – residual AND full • Shooter’s residual framebuffer is simply cleared

  13. Next Shooter Selection • Trivial to select element with most energy left • Multiply lowest mipmap level by area • HOWEVER, two things to consider • Visual impact of color bleeding • Frame-frame coherence

  14. Problem #1 Importance of color over brightness 6 shots over 64x64 (~0.3 seconds)

  15. Problem #2 Note the flickering as the light source descends, which can be pretty distracting

  16. Problem #2 (2)

  17. Different Preference Measures • Sum of components • Max of components • Variance (+mean) • Luminance

  18. SO: • Radiosity at realtimeframerates (1-10 fps) • Quality tradeoffs for speed: • Size of hemisphere texture • Subdivision size • Aggregate shooting • Preferential next shooter selection • Brightness • Color • Coherence

More Related