ݺߣ

ݺߣShare a Scribd company logo

Dynamic lighting in GOW3

?Download as PPTX, PDF?
?
Chung-Yuan Lee
Chung-Yuan Lee

This document describes a lighting method for console games that addresses bottlenecks in graphics processing. It presents a hybrid vertex lights approach that calculates lighting per vertex by combining multiple lights into a single aggregate light. This improves performance by reducing the number of lights that need to be processed per pixel. The method works by calculating light weights based on distance, combining light positions and colors, and interpolating the aggregate light across vertices.

1 of 40
A LIGHTING METHOD FOR CONSOLE GAMES Resource: Dynamic lighting in GOW3 Vassily Filippov (Sony Santa Monica) Advances in Real-Time Rendering in 3D Graphics and Games, SIGGRAPH 2011
INTRODUCTION ? Three major issues for console games ? Modeling Mesh Color ? Lighting Today?s emphasis ? Animation ? PS3?s graphics processing is handled by the NVIDIA RSX 'Reality Synthesizer? ? RSX ? Nvidia GeForce 7800 Architecture ? 256 MB GDDR3 RAM at 700 MHz ? RSX is a bottleneck
TRADITIONAL METHOD ? ? (N Li )li ? ? i ( N Li 0 ) ? Li : Light direction li : Luminance Light Light Light Light
THIS METHOD ? ? (N L Agg ) l Agg ? L Agg : Aggregate Light direction l Agg : Aggregate Luminance Light
HYBRID VERTEX LIGHTS ? For 1 light identical to pixel lights ? For multiple lights ? Calculate distance falloff per vertex. ? Combine into a single Aggregate Light per- vertex ? Interpolate Aggregate Light Position per pixel
INTERPOLATING DIRECTION Light
INTERPOLATING DIRECTION Light A L=(?LA+?LB) B
INTERPOLATING DIRECTION WRONG A L=normalize(?LA+?LB) B
INTERPOLATING POSITION Light A B
INTERPOLATING POSITION Light A B L = Light_Pos?(?A+?B)
INTERPOLATING POSITION BETTER Light A B Normalize(Light_Pos?(?A+?B))
BUT STILL NOT SO GOOD Ideal Aggregate aggregate light light Light 1 Light 0 Vertex V ? Light with falloff in real world, but light weights will both set to 1 here ? Resulting lighting will over-emphasize Light 1
BETTER APPROACH Light 1 Light 0 W0 = 1/L0 L0 L1 W1 = 1/L1 Vertex V ? Step1 - Subtract the world vertex position from each world light position to create relative vectors ? Step2 - Calculate lengths And weights(light intensities are 1 for both lights)
BETTER APPROACH Light 1 Light 0 W0 = 1/L0 W1 = 1/L1 Vertex V ? Step3 - multiply relative vectors by weights to go to direction domain ? Step4 - Add up light directions
BETTER APPROACH Light 1 Light 0 W0 = 1/L0 Vertex V Wtotal + W1 = 1/L1 ? Step5 - Accumulate weights ? Step6 - Multiply aggregate direction by accumulated weight to go back to positions domain ? We ended up with relative light vector for aggregate light
BETTER APPROACH Light 1 Light 0 Vertex V ? Step7 - Add vertex world position to it to get world position of the aggregate light
BETTER APPROACH Light 0 ( Li V ) Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
AGGREGATE LIGHT COLOR ? Calculate aggregate light position ? Calculate normalized light directions ? Calculate dot products x + x =
AGGREGATE LIGHT COLOR ? ? C agg Ci (( L i L agg )) i [ 0 .. n ] Cagg is aggregate light color at the Vertex V Ci is color of Light i Li is direction to Light i at Vertex V Lagg is direction to aggregate light
AGGREGATE LIGHT COLOR ? ? C agg Ci (( L i L agg )) i [ 0 .. n ] Cagg is aggregate light color at the Vertex V Ci is color of Light i x + x = Li is direction to Light i at Vertex V Lagg is direction to aggregate light How well does this approximate actual lighting by several lights?
WHY THIS WORKS ? This is what is tried to approximate ? ? ? ? (N Li )ci i ( N Li 0 ) ? Assume lights have a single scalar luminance value and no color. ? ? ? ? (N Li )li i ( N Li 0)
WHY THIS WORKS Aggregate light approximation (with luminances) ? ? ? ? (N Li ) li (N L Agg ) l Agg ? ? i ( N Li 0) Our approximation Exact luminance calculation
WHY THIS WORKS ? ? ? ? ? ? (N Li )li N ( l i Li ) i ( N Li 0) ? X ? ? l i Li X
WHY THIS WORKS Light 0 ( Lposi V ) Wi i [ 0.. n ] Lposi V Lposagg V Vertex V Wi i [ 0.. n ] Lposi V li
WHY THIS WORKS Light 0 ( Lposi V ) li i [ 0.. n ] Lposi V Lposagg V Vertex V li i [ 0.. n ] Lposi V Lrelagg
WHY THIS WORKS Light 0 ( Lposi V ) li i [ 0.. n ] Lposi V Lagg Lrelagg Vertex V li i [ 0.. n ] Lposi V Lagg li ( Lposi V) i [ 0.. n ] Lposi V
WHY THIS WORKS Light 0 ( Lposi V) li i [ 0.. n ] Lposi V Lagg Lagg Vertex V ( Lposi V) li i [ 0.. n ] Lposi V Li
WHY THIS WORKS Light 0 li Li i [ 0.. n ] Lagg Lagg Vertex V li Li i [ 0.. n ] ? X
WHY THIS WORKS ? ? ? ? ? ? ? (N Li ) li (N L Agg ) l Agg X i ( N Li 0) Lagg ? ? ? X l i Li X ? lagg X
VECTOR ARITHMETIC Adding multiple vectors Length of the result is equal to the sum of projections V2 V1 V0
WHY THIS WORKS C THE LAST STEP ? l Agg X
WHY THIS WORKS C THE LAST STEP ? ? l Agg X li Li i n
WHY THIS WORKS C THE LAST STEP ? l j Lj ? ? ? j n l Agg X li Li li Li i n i n ? l j Lj j n LAgg
WHY THIS WORKS C THE LAST STEP ? ? l Agg li ( Li LAgg ) i For luminance our approximation is exact Extending to RGB: ? ? C Agg Ci ( Li LAgg ) Which is what this method use
RESULT God Of War III Runs Natively At 720p With 1080p Upscaling, and stable 60fps most of the time

More Related Content

Dynamic lighting in GOW3

  • 1. A LIGHTING METHOD FOR CONSOLE GAMES Resource: Dynamic lighting in GOW3 Vassily Filippov (Sony Santa Monica) Advances in Real-Time Rendering in 3D Graphics and Games, SIGGRAPH 2011
  • 2. INTRODUCTION ? Three major issues for console games ? Modeling Mesh Color ? Lighting Today?s emphasis ? Animation ? PS3?s graphics processing is handled by the NVIDIA RSX 'Reality Synthesizer? ? RSX ? Nvidia GeForce 7800 Architecture ? 256 MB GDDR3 RAM at 700 MHz ? RSX is a bottleneck
  • 3. TRADITIONAL METHOD ? ? (N Li )li ? ? i ( N Li 0 ) ? Li : Light direction li : Luminance Light Light Light Light
  • 4. THIS METHOD ? ? (N L Agg ) l Agg ? L Agg : Aggregate Light direction l Agg : Aggregate Luminance Light
  • 5. HYBRID VERTEX LIGHTS ? For 1 light identical to pixel lights ? For multiple lights ? Calculate distance falloff per vertex. ? Combine into a single Aggregate Light per- vertex ? Interpolate Aggregate Light Position per pixel
  • 7. INTERPOLATING DIRECTION Light A L=(?LA+?LB) B
  • 8. INTERPOLATING DIRECTION WRONG A L=normalize(?LA+?LB) B
  • 10. INTERPOLATING POSITION Light A B L = Light_Pos?(?A+?B)
  • 11. INTERPOLATING POSITION BETTER Light A B Normalize(Light_Pos?(?A+?B))
  • 12. BUT STILL NOT SO GOOD Ideal Aggregate aggregate light light Light 1 Light 0 Vertex V ? Light with falloff in real world, but light weights will both set to 1 here ? Resulting lighting will over-emphasize Light 1
  • 13. BETTER APPROACH Light 1 Light 0 W0 = 1/L0 L0 L1 W1 = 1/L1 Vertex V ? Step1 - Subtract the world vertex position from each world light position to create relative vectors ? Step2 - Calculate lengths And weights(light intensities are 1 for both lights)
  • 14. BETTER APPROACH Light 1 Light 0 W0 = 1/L0 W1 = 1/L1 Vertex V ? Step3 - multiply relative vectors by weights to go to direction domain ? Step4 - Add up light directions
  • 15. BETTER APPROACH Light 1 Light 0 W0 = 1/L0 Vertex V Wtotal + W1 = 1/L1 ? Step5 - Accumulate weights ? Step6 - Multiply aggregate direction by accumulated weight to go back to positions domain ? We ended up with relative light vector for aggregate light
  • 16. BETTER APPROACH Light 1 Light 0 Vertex V ? Step7 - Add vertex world position to it to get world position of the aggregate light
  • 17. BETTER APPROACH Light 0 ( Li V ) Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
  • 18. BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
  • 19. BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
  • 20. BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
  • 21. BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
  • 22. BETTER APPROACH ( Li V ) Light 0 Wi i [ 0.. n ] Li V Lagg V Wi Vertex V i [ 0.. n ] Li V Lagg is aggregate light position Li is light position for Light i Wi is light weight for Light i based on intensity of Light i at V n is number of lights in the current light context V is position of the vertex for which we combine lights
  • 23. AGGREGATE LIGHT COLOR ? Calculate aggregate light position ? Calculate normalized light directions ? Calculate dot products x + x =
  • 24. AGGREGATE LIGHT COLOR ? ? C agg Ci (( L i L agg )) i [ 0 .. n ] Cagg is aggregate light color at the Vertex V Ci is color of Light i Li is direction to Light i at Vertex V Lagg is direction to aggregate light
  • 25. AGGREGATE LIGHT COLOR ? ? C agg Ci (( L i L agg )) i [ 0 .. n ] Cagg is aggregate light color at the Vertex V Ci is color of Light i x + x = Li is direction to Light i at Vertex V Lagg is direction to aggregate light How well does this approximate actual lighting by several lights?
  • 26. WHY THIS WORKS ? This is what is tried to approximate ? ? ? ? (N Li )ci i ( N Li 0 ) ? Assume lights have a single scalar luminance value and no color. ? ? ? ? (N Li )li i ( N Li 0)
  • 27. WHY THIS WORKS Aggregate light approximation (with luminances) ? ? ? ? (N Li ) li (N L Agg ) l Agg ? ? i ( N Li 0) Our approximation Exact luminance calculation
  • 28. WHY THIS WORKS ? ? ? ? ? ? (N Li )li N ( l i Li ) i ( N Li 0) ? X ? ? l i Li X
  • 29. WHY THIS WORKS Light 0 ( Lposi V ) Wi i [ 0.. n ] Lposi V Lposagg V Vertex V Wi i [ 0.. n ] Lposi V li
  • 30. WHY THIS WORKS Light 0 ( Lposi V ) li i [ 0.. n ] Lposi V Lposagg V Vertex V li i [ 0.. n ] Lposi V Lrelagg
  • 31. WHY THIS WORKS Light 0 ( Lposi V ) li i [ 0.. n ] Lposi V Lagg Lrelagg Vertex V li i [ 0.. n ] Lposi V Lagg li ( Lposi V) i [ 0.. n ] Lposi V
  • 32. WHY THIS WORKS Light 0 ( Lposi V) li i [ 0.. n ] Lposi V Lagg Lagg Vertex V ( Lposi V) li i [ 0.. n ] Lposi V Li
  • 33. WHY THIS WORKS Light 0 li Li i [ 0.. n ] Lagg Lagg Vertex V li Li i [ 0.. n ] ? X
  • 34. WHY THIS WORKS ? ? ? ? ? ? ? (N Li ) li (N L Agg ) l Agg X i ( N Li 0) Lagg ? ? ? X l i Li X ? lagg X
  • 35. VECTOR ARITHMETIC Adding multiple vectors Length of the result is equal to the sum of projections V2 V1 V0
  • 36. WHY THIS WORKS C THE LAST STEP ? l Agg X
  • 37. WHY THIS WORKS C THE LAST STEP ? ? l Agg X li Li i n
  • 38. WHY THIS WORKS C THE LAST STEP ? l j Lj ? ? ? j n l Agg X li Li li Li i n i n ? l j Lj j n LAgg
  • 39. WHY THIS WORKS C THE LAST STEP ? ? l Agg li ( Li LAgg ) i For luminance our approximation is exact Extending to RGB: ? ? C Agg Ci ( Li LAgg ) Which is what this method use
  • 40. RESULT God Of War III Runs Natively At 720p With 1080p Upscaling, and stable 60fps most of the time