CS 4363/6353 BASIC RENDERING

THE GRAPHICS PIPELINE OVERVIEW Vertex Processing Coordinate transformations Compute color for each vertex Clipping and Primitive Assembly Assemble sets of vertices into lines and polygons Clipping volume culls out geometry outside, and clips geo that straddles Rasterization Determine which pixels are inside each polygon (primitive) The output is a set of fragments for each primitive Fragment Processing Fills in the pixels in the frame buffer (what you’re seeing right now!)

SHADERS (FROM 100,000 FEET. WELL, MAYBE 10,000 FEET) Shaders: Expose the programmable pipeline ! Allow us to manipulate vertices and pixel colors Look very similar to C void main() Must be compiled and linked from source Generally have two shaders: Vertex shader (to handle vertex transformations and lighting) Fragment shaders (to handle per-pixel operations like lighting to determine color) Hundreds (or thousands) of GPUs are available

SHADERS AND PASSING DATA Must feed shaders some data! Three ways to pass. Attributes (vertex shaders only): data that changes per vertex A four component vector (regardless if you use it or not) Copied from your OpenGL program into a buffer Uniforms (vertex/fragment shaders): A single value that is shared for all attributes Common for transformation matrices (vertex shader) Texture (mostly fragment shaders): Used for texture data

SIMPLE VERTEX EXAMPLE #version 150 in vec4 vPosition; // This data is from your OpenGL code void main () { gl_Position = vPosition; }

SIMPLE FRAGMENT SHADER #version 150 out vec4 fColor; void main () { fColor = vec4(1.0, 0.0, 0.0, 1.0); } // Hard-coded red!

SHADERS AND PASSING DATA outs/ins Used for passing data between shaders Client (OpenGL code) has no access to these variables Vertex shader’s out variable corresponds to the fragment shader’s in variable.

RENDERING OPTIONS We typically create a single batch of vertices to draw We generally set up all “features” before drawing anything State machine mentality We always have vertices, but we can render them in 7 different ways GL_POINTS GL_LINES GL_LINE_STRIP GL_LINE_LOOP GL_TRIANGLES** GL_TRIANGLE_STRIP GL_TRIANGLE_FAN

RENDERING OPTIONS ( GL_POINTS ) Can change the point sizes, but not important right now Points are always square unless anti-aliased

RENDERING OPTIONS ( GL_LINES ) Connects in pairs (line segments), so should have an even number of points Change line width with glLineWidth (GLfloat width);

RENDERING OPTIONS ( GL_LINE_STRIP ) In a connect-the-dots fashion, draw from one vertex to another How would you do this with GL_LINES ?

RENDERING OPTIONS ( GL_LINE_LOOP ) Closes the loop Is typically what you would use for outlines/tracing

RENDERING OPTIONS ( OTHERS ) Some of these are allowed…

RENDERING OPTIONS ( OTHERS ) Some of these aren’t… NO QUADS! Triangles only…

TRIANGLE WINDING Simply means the order of the vertices you specify Clockwise Counter-Clockwise Why is this important? Clockwise is back facing Counter-clockwise is front facing Long story short: If you specify in reverse order, sometimes you won’t see anything or it will be reversed Can reverse using glFrontFace (GL_CW); V0V0 V1V1 V2V2 V0V0 V2V2 V1V1

TRIANGLE STRIPS Specify the first triangle (V 0, V 1, V 2 ) V0V0 V1V1 V2V2

TRIANGLE STRIPS Specify the first triangle (V 0, V 1, V 2 ) The next vertex (V 3 ) creates a new triangle (V 1, V 2, V 3 ) V0V0 V1V1 V2V2 V3V3

TRIANGLE STRIPS Specify the first triangle (V 0, V 1, V 2 ) The next vertex (V 3 ) creates a new triangle (V 1, V 2, V 3 ) The next vertex (V 4 ) creates a new triangle (V 2, V 3, V 4 ) V0V0 V1V1 V2V2 V3V3 V4V4

EXAMPLE From Wikipedia…

TRIANGLE FANS Can create a fan where V 0 is the central vertex Specify first triangle, then each new vertex is a wedge of the fan V0V0 V1V1 V2V2

TRIANGLE FANS Can create a fan where V 0 is the central vertex Specify first triangle, then each new vertex is a wedge of the fan Still uses V 0 V0V0 V1V1 V2V2 V3V3

TRIANGLE FANS Can create a fan where V 0 is the central vertex Specify first triangle, then each new vertex is a wedge of the fan Still uses V 0 V0V0 V1V1 V2V2 V3V3 V4V4

EXAMPLE OF TRIANGLE FAN

CULLING AND DEPTH TESTING You’re going to be drawing a lot of triangles What happens if you draw one triangle on top of another? What if the second triangle is far away? Sort triangles of an object (painter’s algorithm)? What about several objects on the screen? Also, should you be able to see the inside of geometry? Basically, there are two problems: Unseen triangles are unlit The depth of the triangles is important

HOUSTON, WE HAVE A PROBLEM (IMAGE FROM OPENGL SUPERBIBLE) Sometimes rendering the far-side triangles (which are unlit)

BACKFACE CULLING Simply means “Don’t draw triangles that don’t face the camera” Two steps in OpenGL glEnable (GL_CULL_FACE); glCullFace (GL_BACK); glCullFace could also use: GL_FRONT GL_FRONT_AND_BACK

CULLING THE BACKFACE(S) Problem solved?

BUT WAIT! (IMAGES FROM OPENGL SUPERBIBLE)

DEPTH TESTING Just because we culled the back-facing triangles doesn’t mean they’re sorted! Depth testing: Removes hidden surfaces Each pixel has a depth (z-value) Higher values mean closer to the camera This value is stored in the depth buffer glEnable (GL_DEPTH_TEST); Is this starting to make sense? glutInitDisplayMode (GLUT_DOUBLE|GLUT_RGBA|GLUT_DEPTH);

POLYGON RENDERING MODES Polygons do not have to be filled We have 3 modes to draw polygons: GL_FILL – what we’ve been using GL_LINE – 3D wireframe GL_POINT – just the vertices Call glPolygonMode() to change rendering value: // renders front and back facing polys in wireframe glPolygonMode (GL_FRONT_AND_BACK, GL_LINE);

GL_LINE Backface culling and depth testing are turned on

A NOTE ABOUT POLYGON OFFSET You can skip it Sometimes draw two triangle very close to the same depth (called decaling) This creates “z-fighting” Part of the further polygon shows Before rendering closer triangle: glEnable (GL_POLYGON_OFFSET_LINE); glPolygonOffset(-1.0f, -1.0f);

SCISSOR TEST We won’t be using this in our code, but: Used to increase performance Updates only the portion within a defined area (i.e. doesn’t update anything outside of that area) By default, scissor test is the size of the window Use glScissor (int x, int y, int width, int height) : glEnable (GL_SCISSOR_TEST); glScissor (100, 100, 600, 400); // only render in that area

BLENDING Without depth testing, color values overwrite one another With depth testing, new fragments may replace old ones Discards further fragments This no longer happens with OpenGL blending: glEnable (GL_BLENDING); Remember, each color has a red, green, blue and alpha!

SPECIFYING HOW TO BLEND We must specify how the blending occurs Destination color is the color already in the color buffer Source color is the one we’re about to write into the color buffer C f = (C s * S) + (C d * D) Where: C f is the final color, C s is the source color and C d is the destination color S is the source blending factor D is the destination blending factor

MOST COMMON METHOD glBlendFunc (GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA); Take source and multiply rgb (colors) by its alpha value Take destination and multiply rgb by (1-source alpha) Example: C d = (1.0f, 0.0f, 0.0f, 1.0f) //Red C s = (0.0f, 0.0f, 1.0f, 0.6f) // Blue, with source alpha 0.6 D = 1.0f – 0.6 == 0.4f C f = (Blue*0.6) + (Red*0.4)

EXAMPLE

ONE FINAL NOTE We can change the underlying equation as well using glBlendEquation(): GL_FUNC_ADD C f = (C s *S)+(C d *D) GL_FUNC_SUBTRACT C f = (C s *S)-(C d *D) GL_FUNC_REVERSE_SUBTRACT C f = (C d *D) - (C s *S) GL_MIN C f = min(C s,C d ) GL_MAX C f = max(C s,C d )

ANTIALIASING We have square pixels, which make the image look computer-generated The visual aspect of this is called “the jaggies” To eliminate, OpenGL uses blending of source with surrounding destination pixels So: glEnable (GL_BLEND); glBlendFunc (GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA); Then: glEnable (GL_POINT_SMOOTH); // and/or glEnable (GL_LINE_SMOOTH); // and/or glEnable (GL_POYGON_SMOOTH);// ** // ** outdated or not supported at all!

EXAMPLE

MULTISAMPLING Helps to smooth out polygons Creates another buffer (color, depth and stencil) All primitives are sampled multiple times, then “averaged” You take a performance hit here, but looks good! Point/Line antialiasing is disable when multisampling is enabled First: glutInitDisplayMode (GLUT_DOUBLE| GLUT_RGB| GLUT_DEPTH| GLUT_MULTISAMPLE); Then: glEnable (GL_MULTISAMPLE);