1. Shadow Mapping Chun-Fa Chang National Taiwan Normal University

2. Advanced Texture Mapping  Using multiple textures  Multi-Pass textures 1 st Pass: render the scenes as usual Create textures from the output images 2 nd Pass: render the scenes again using the created texture

3. Using Textures in GLSL Shader  sampler2D data type in GLSL  Binding to the C/C++ program through glGetUniformLocation()  See the myTexture variable in Lab 7 in both the fragment shader and, the C code setShaders().

4. Shadow Map  Using two textures: color and depth  Relatively straightforward design using pixel (fragment) shaders on GPUs.

5. Image Source: Cass Everitt et al., “Hardware Shadow Mapping” NVIDIA SDK White PaperHardware Shadow Mapping Eye’s ViewLight’s ViewDepth/Shadow Map

6. Basic Steps of Shadow Maps 1. Render the scene from the light’s point of view, 2. Use the light’s depth buffer as a texture (shadow map), 3. Projectively texture the shadow map onto the scene,  Use “TexGen” or shader 4. Use “texture color” (comparison result) in fragment shading.

7. What’re in the Example Code?  A C++ class for storing matrix state: class OpenGL_Matrix_State { void Save_Matrix_State(); void Restore_Matrix_State(); void Set_Texture_Matrix(); }  A proxy rectangle for debug

8. (1) Rendering from Light’s View  Set the camera to the light position.  Viewport set to the same size as the texture.  To avoid the floating point precision problem (casting its own shadow to a surface), depth must be shifted: glPolygonOffset(...,...); glEnable(GL_POLYGON_OFFSET_FILL);  Shading could be turned off We only care about the depth!

9. (2) Creation of Shadow Map (Texture)  Draw the objects (from light’s view)  To create a depth texture, use: glTexImage2D( GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT, shadowMapSize, shadowMapSize, 0, GL_DEPTH_COMPONENT, GL_UNSIGNED_BYTE, 0);  Then use glCopyTexSubImage2D() to copy the frame buffer to the depth texture.

10. (3) Generation of Texture Coordinates  When we render the scene again from the normal camera view: We store the light’s view to the texture matrix. The texture matrix is then passed to the GLSL shaders. gl_TextureMatrix[0] * vertex gives us the homogeneous coordinates in light space Divide by w to obtain the texture coordinates. Watch out! Must shift from [-1, 1] to [0,1]

11. Normalized Coordinates  Independent of the screen resolution or window size.  Clip coordinates: after Model-View and Projection transformation.  Normalized Device Coordinates (NDC): after division by w.

12. (4) Depth Comparison in Fragment Shader  Compare two depths: Depth read from the shadow map Depth by transformation to the light space  In the shadow if ____?_(your exercise)____  Set a darker color for shadowed surfaces

13. More GPU Programming and GPGPU Chun-Fa Chang National Taiwan Normal University

14. Calculator vs. Computer  What is the difference between a calculator and a computer?  Doesn ’ t a compute-r just “ compute ” ?  The Casio fx3600p calculated can be programmed (38 steps allowed).

15. Turing Machine  Can be adapted to simulates the logic of any computer that could possibly be constructed.  von Neumann architecture implements a universal Turing machine.  Look them up at Wikipedia!

17. Simplified View  The Data Flow: 3D Polygons (+Colors, Lights, Normals, Texture Coordinates … etc.)  2D Polygons  2D Pixels (I.e., Output Images) Transform (& Lighting) Rasterization

18. Global Effects translucent surface shadow multiple reflection

19. Local vs. Global

20. How Does GPU Draw This?

21. Quiz  Q1: A straightforward GPU pipeline give us local illumination only. Why?  Q2: What typical effects are missing? Hint: How is an object drawn? Do they consider the relationship with other objects? Shadow, reflection, and refraction…

22.  Wait but I ’ ve seen shadow and reflection in games before … With ShadowsWithout Shadows

23. Faked Global Illumination  Shadow, Reflection, BRDF … etc.  In theory, real global illumination is not possible in current graphics pipeline: Conceptually a loop of individual polygons. No interaction between polygons.  Can this be changed by multi-pass rendering?

24. Case Study: Shadow Map  Using two textures: color and depth  Relatively straightforward design using pixel (fragment) shaders on GPUs.

25. Adding “ Memory ” to the GPU Computation  Modern GPUs allow: The usage of multiple textures. Rendering algorithms that use multiple passes. Transform (& Lighting) Rasterization Textures