Dr. Scott Schaefer

1
Programmable Shaders

• Dr. Scott Schaefer

2
Graphics Cards Performance

• Nvidia Geforce 6800 GTX1
• 6.4 billion pixels/sec
• Nvidia Geforce 7900 GTX2
• 15.6 billion pixels/sec
• Xbox 3603
• 16 billion pixels/sec (4X AA)
• Nvidia Geforce 8800 GTX4
• 36.8 billion pixels/sec
• Nvidia Geforce GTX 2805
• 48.2 billion pixels/sec

1 http//www.nvidia.com/page/geforce_6800.html
2 http//www.nvidia.com/page/geforce_7900.html
3 http//news.com.com/Xboxspecsrevealed/2100-10
43_3-5705372.html 4 http//www.nvidia.com/page/
geforce_8800.html 5 http//www.nvidia.com/object
/geforce_gtx_280.html
3
Parallel Processing Power

• Nvidia Geforce GTX 280
• 240 programmable processors
• 1.3 GHz each
• 141.7 GB/s memory bandwidth
• 1GB memory

http//www.nvidia.com/object/geforce_gtx_280.html
4
Parallel Processing Power
IBMs ASCI Red, 1.338 TFLOPS Fastest Computer in
the World 1997
ATIs Radeon HD 4850, 1.2 TFLOPS
5
Graphics Pipeline
Vertices
6
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
7
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
Transformed Vertices
8
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
Transformed Vertices
Viewport Transformation
9
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
Vertex Index Stream
Transformed Vertices
Viewport Transformation
Triangle Setup
10
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
Vertex Index Stream
Transformed Vertices
Viewport Transformation
Triangle Setup
Backface Culling
11
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
Vertex Index Stream
Transformed Vertices
Viewport Transformation
Triangle Setup
Backface Culling
Clipping
12
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
Vertex Index Stream
Transformed Vertices
Viewport Transformation
Triangle Setup
Backface Culling
Clipping
Interpolation/Rasterization
13
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
Vertex Index Stream
Transformed Vertices
Viewport Transformation
Triangle Setup
Backface Culling
Clipping
Interpolation/Rasterization
Pixel Location/Color/Depth
14
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
Vertex Index Stream
Transformed Vertices
Viewport Transformation
Triangle Setup
Backface Culling
Clipping
Interpolation/Rasterization
Pixel Location/Color/Depth
Visibility Determination
15
Graphics Pipeline
Vertices
Vertex Transformation/Lighting
Vertex Index Stream
Transformed Vertices
Viewport Transformation
Triangle Setup
Backface Culling
Clipping
Interpolation/Rasterization
Pixel Location/Color/Depth
Visibility Determination
Frame Buffer
16
Programmable Graphics Pipeline
Vertices
Vertex Shader
Vertex Index Stream
Transformed Vertices/ Normals/Texture coords/
Viewport Transformation
Triangle Setup
Backface Culling
Clipping
Interpolated Vertex Data
Interpolation/Rasterization
Pixel Shader
Pixel Location
Visibility Determination
Color/Depth
Frame Buffer
17
Shader Programming

• Many different languages
• Assembly
• OpenGL Shading Language
• Nvidias CG
• Microsofts HLSL
• Different capabilities based on shader model
• Register count
• Instructions
• Maximum number of instructions

18
Vertex Shaders

• Input anything associated with vertices
• Position, normal, texture coordinates, etc
• Output transformed vertices
• MUST output position
• Can produce color, normal, texture coordinates,
  etc

19
Vertex Shaders

• // vertex shader output structure
• struct VS_OUTPUT
• float4 Pos POSITION

20
Vertex Shaders

• VS_OUTPUT VS(
• float3 InPos POSITION // Vertex position
  in model space
  
• )
• VS_OUTPUT Out (VS_OUTPUT)0
• // transform the position
• float3 transformedPos mul(float4(InPos, 1),
  (float4x3)World)
  
• Out.Pos mul(float4(transformedPos,1),
  ViewProjection)
  
• return Out

21
Pixel Shaders

• Input Vertex data produced from vertex shader
• Output
• MUST output color
• Can output depth as well
• Cannot change location of pixel on screen

22
Pixel Shaders

• float4 PS ( VS_OUTPUT In ) COLOR
• // may perform texture lookup, depth effects,
  fog, etc
  
• return float4 ( 1, 1, 1, 1 )

23
Gouraud Shading Example

• // vertex shader output structure
• struct VS_OUTPUT
• float4 Pos POSITION
• float4 Color COLOR

24
Gouraud Shading Example

• VS_OUTPUT VS(
• float3 InPos POSITION, // Vertex
  position in model space
  
• float3 InNormal NORMAL // Vertex normal in
  model space
  
• )
• VS_OUTPUT Out (VS_OUTPUT)0
• // transform the position and normal
• float3 transformedPos mul(float4(InPos, 1),
  (float4x3)World)
  
• Out.Pos mul(float4(transformedPos,1),
  ViewProjection)
  
• float3 transNormal mul(InNormal,
  (float3x3)World) // normal (view space)
  
• Out.Color float4 ( calcColor ( normalize (
  lightPos transformedPos ), transNormal,
  
• 
  normalize ( eyePos transformedPos ) ), 1 )
  
• return Out

25
Gouraud Shading Example

• float3 calcColor ( float3 lightVec, float3
  normal, float3 eyeToVertex )
  
• float3 color 0
• color lightColor MaterialAmbient
• color lightColor MaterialDiffuse max
  ( 0, dot ( normal, lightVec ) )
  
• float3 R normalize ( reflect ( lightVec,
  normal ) )
  
• color lightColor MaterialSpecular pow
  ( max ( 0, dot ( R,
  
• eyeToVertex ) ),
  MaterialSpecularPower )
  
• return color

26
Gouraud Shading Example

• float4 PS ( VS_OUTPUT In ) COLOR
• return In.Color

27
Phong Shading Example

• // vertex shader output structure
• struct VS_OUTPUT
• float4 Pos POSITION
• float3 Normal TEXCOORD0
• float3 TransformedPos TEXCOORD1

28
Phong Shading Example

• VS_OUTPUT VS(
• float3 InPos POSITION, // Vertex
  position in model space
  
• float3 InNormal NORMAL // Vertex normal in
  model space
  
• )
• VS_OUTPUT Out (VS_OUTPUT)0
• // transform the position and normal
• Out.TransformedPos mul(float4(InPos, 1),
  (float4x3)World)
  
• Out.Pos mul(float4(Out.TransformedPos,1),
  ViewProjection)
  
• Out.Normal mul(InNormal, (float3x3)World)
  // normal (view space)
  
• return Out

29
Phong Shading Example

• float4 PS ( VS_OUTPUT In ) COLOR
• // vector from vertex towards eye
• float3 EyeToVertex normalize (
  In.TransformedPos - EyePos )
  
• float3 normal normalize ( In.Normal )
• float4 color calcColor ( normalize (
  lightPos In.TransformedPos ), normal,
  
• 
  EyeToVertex )
  
• return color

30
General Purpose GPU Programming

• Limited success because problems must be crammed
  into graphics pipeline
  
• General purpose computation starting to become
  available
  
• Nvidias CUDA
• Peakstream