Mohammad Shaker

mohammadshaker.com

@ZGTRShaker

2014

OpenGL Graphics

L06- PERFORMANCE

Performance

How can you up your OpenGL app performance?

• Different Techniques:

– Vertex buffer and vertex arrays

– Display Lists

– Frustum culling

– Other methods

Vertex Buffer Object (VBO)Vertex Array Object (VAO)

Vertex and Array Buffers

• Look at this,

Vertex and Array Buffers

Good oder Bad?

Vertex and Array Buffers

Good oder Bad?

Vertex and Array Buffers

Vertex and Array Buffers

Vertex and Array Buffers

huge performance benefit!

VBO and VAOExample: Creating Cube Using VBO and VAO

Initializing the Cube’s Data

• We’ll build each cube face from individual triangles

• We need to determine how much storage is required:

– (6 faces)(2 triangles/face)(3 vertices/triangle)

const int NumVertices = 36;

• To simplify communicating with GLSL, we’ll use a vec4 class (implemented in

C++) similar to GLSL’s vec4 type

– We’ll also typedef it to add logical meaning:

typedef vec4 point4;typedef vec4 color4;

Initializing the Cube’s Data

• Our cube has two attributes per vertex:

– position

– color

• We create two arrays to hold the VBO data:

point4 points[NumVertices];color4 colors[NumVertices];

Initializing the Cube’s Data

// Vertices of a unit cube centered at origin, sides aligned with axes

point4 vertex_positions[8] = {

point4(-0.5, -0.5, 0.5, 1.0 ),

point4(-0.5, 0.5, 0.5, 1.0 ),

point4( 0.5, 0.5, 0.5, 1.0 ),

point4( 0.5, -0.5, 0.5, 1.0 ),

point4(-0.5, -0.5, -0.5, 1.0 ),

point4(-0.5, 0.5, -0.5, 1.0 ),

point4( 0.5, 0.5, -0.5, 1.0 ),

point4( 0.5, -0.5, -0.5, 1.0 )

};

// RGBA colors

color4 vertex_colors[8] = {

color4(0.0, 0.0, 0.0, 1.0 ), // black

color4(1.0, 0.0, 0.0, 1.0 ), // red

color4(1.0, 1.0, 0.0, 1.0 ), // yellow

color4(0.0, 1.0, 0.0, 1.0 ), // green

color4(0.0, 0.0, 1.0, 1.0 ), // blue

color4(1.0, 0.0, 1.0, 1.0 ), // magenta

color4(1.0, 1.0, 1.0, 1.0 ), // white

color4(0.0, 1.0, 1.0, 1.0 ) // cyan

};

Initializing the Cube’s Data

// quad() generates two triangles for each face and assigns colors to the vertices

int Index = 0; // global variable indexing into VBO arrays

void quad(int a, int b, int c, int d )

{

colors[Index] = vertex_colors[a]; points[Index] = vertex_positions[a]; Index++;

colors[Index] = vertex_colors[b]; points[Index] = vertex_positions[b]; Index++;

colors[Index] = vertex_colors[c]; points[Index] = vertex_positions[c]; Index++;

colors[Index] = vertex_colors[a]; points[Index] = vertex_positions[a]; Index++;

colors[Index] = vertex_colors[c]; points[Index] = vertex_positions[c]; Index++;

colors[Index] = vertex_colors[d]; points[Index] = vertex_positions[d]; Index++;

}

// generate 12 triangles: 36 vertices and 36 colors

Void colorcube()

{

quad(1, 0, 3, 2);

quad(2, 3, 7, 6);

quad(3, 0, 4, 7);

quad(6, 5, 1, 2);

quad(4, 5, 6, 7);

quad(5, 4, 0, 1);

}

Creating the VAO

• VAOs store the data of a geometric object

• Steps for using a VAO

1. Generate VAO names by calling glgenvertexarrays()

2. Bind a specific VAO for initialization by calling glbindvertexarray()

3. Update VBOs associated with this VAO

4. Bind VAO for use in rendering

• This approach allows a single function call to specify all the data for an objects

– previously, you might have needed to make many calls to make all the data current

GLuint vao; // Create a vertex array object

glGenVertexArrays(1, &vao);

glBindVertexArray(vao);

• Vertex data must be stored in a VBO, and associated with a VAO

• The code-flow is similar to configuring a VAO:– generate VBO names by calling

glGenBuffers()

– bind a specific VBO for initialization by calling

glBindBuffer(GL_ARRAY_BUFFER, …)

– load data into VBO using

glBufferData(GL_ARRAY_BUFFER, …)

– bind VAO for use in rendering

glBindVertexArray()

Creating the VBO

GLuint buffer; // Create and initialize a buffer object

glGenBuffers(1, &buffer);

glBindBuffer(GL_ARRAY_BUFFER, buffer);

glBufferData(GL_ARRAY_BUFFER, sizeof(points) + sizeof(colors), NULL, GL_STATIC_DRAW);

glBufferSubData(GL_ARRAY_BUFFER, 0, sizeof(points), points);

glBufferSubData(GL_ARRAY_BUFFER, sizeof(points), sizeof(colors), colors);

• Vertex data must be stored in a VBO, and associated with a VAO

• The code-flow is similar to configuring a VAO:– generate VBO names by calling

glGenBuffers()

– bind a specific VBO for initialization by calling

glBindBuffer(GL_ARRAY_BUFFER, …)

– load data into VBO using

glBufferData(GL_ARRAY_BUFFER, …)

– bind VAO for use in rendering

glBindVertexArray()

Creating the VBO

GLuint buffer; // Create and initialize a buffer object

glGenBuffers(1, &buffer);

glBindBuffer(GL_ARRAY_BUFFER, buffer);

glBufferData(GL_ARRAY_BUFFER, sizeof(points) + sizeof(colors), NULL, GL_STATIC_DRAW);

glBufferSubData(GL_ARRAY_BUFFER, 0, sizeof(points), points);

glBufferSubData(GL_ARRAY_BUFFER, sizeof(points), sizeof(colors), colors);

Display ListDisplay lists allow you to define the geometry and execute them multiple times. If you plan to redraw the same geometry multiple times, this is a great way to

save performance!

Immediate Mode versus Display Listed Rendering

• Immediate Mode Graphics

– Primitives are sent to pipeline and display right away

– No memory of graphical entities

• Display Listed Graphics

– Primitives placed in display lists

– Display lists kept on graphics server

– Can be redisplayed with different state

– Can be shared among OpenGL graphics contexts

OpenGL Architecture

DisplayList

PolynomialEvaluator

Per VertexOperations &

PrimitiveAssembly

RasterizationPer Fragment

OperationsFrameBuffer

TextureMemory

CPU

PixelOperations

OpenGL Architecture

DisplayList

PolynomialEvaluator

Per VertexOperations &

PrimitiveAssembly

RasterizationPer Fragment

OperationsFrameBuffer

TextureMemory

CPU

PixelOperations

OpenGL Architecture

DisplayList

PolynomialEvaluator

Per VertexOperations &

PrimitiveAssembly

RasterizationPer Fragment

OperationsFrameBuffer

TextureMemory

CPU

PixelOperations

Display Lists

• Not all OpenGL routines can be stored in display lists

• State changes persist, even after a display list is finished

• Display lists can call other display lists

• Display lists are not editable, but you can fake it

– make a list (A) which calls other lists (B, C, and D)

– delete and replace B, C, and D, as needed

Display Lists and Hierarchy

• Consider model of a car

– Create display list for chassis

– Create display list for wheel

glNewList(CAR, GL_COMPILE);glCallList(CHASSIS);glTranslatef(…);glCallList(WHEEL);glTranslatef(…);glCallList(WHEEL);

…glEndList();

Display List Examplefor Terrain Drawing

Display List for Terrain Drawing

1. Global Scope: GLuint terrainListId;

2. InitGl():

• Generate a list:

terrainListId = glGenLists(1);

• Call the terrain drawing function that you want OpenGL to perform faster

DrawTerrain();

3. DrawTerrain():

– In the first line of this function, tell OpenGL that each instruction from now on must be located in the display list

that’s bound to terrainListId by writing:

• glNewList(terrainListId, GL_COMPILE);

– Now, write the terrain drawing algorithm

– End display list scope by calling

• glEndList();

4. In DrawGLScene() instead of calling DrawTerrain(), call the terrain list:

– glCallList(terrainListId);

Display List for Terrain Drawing

1. Global Scope: GLuint terrainListId;

2. InitGl():

• Generate a list:

terrainListId = glGenLists(1);

• Call the terrain drawing function that you want OpenGL to perform faster

DrawTerrain();

3. DrawTerrain():

– In the first line of this function, tell OpenGL that each instruction from now on must be located in the display list

that’s bound to terrainListId by writing:

• glNewList(terrainListId, GL_COMPILE);

– Now, write the terrain drawing algorithm

– End display list scope by calling

• glEndList();

4. In DrawGLScene() instead of calling DrawTerrain(), call the terrain list:

– glCallList(terrainListId);

GLuint terrianListId;

initGL()

{ terrianListId = glGenLists(1);

DrawTerrain();

}

DrawTerrain()

{ glNewList(terrianListId, GL_COMPILE);

//Terrain Drawing Algorithm

glEndList();

}

DrawGLSene()

{ //Other Stuff

glCallList(terrianListId);

}

Culling and Clipping

Frustum Culling

• Create a clipping plane

– float plane [] = {A, b, C, D};

• Create a clipping plane with it

– glClipPlane(GL_CLIP_PLANE0, plane);

• Enable the clipping plane

– glEnable(GL_CLIP_PLANE0);

Scissor Box

• Additional Clipping Test

glScissor(x, y, w, h )

• any fragments outside of box are clipped

• useful for updating a small section of a viewport

– affects glClear() operations

General Memory Managementhttp://docs.unity3d.com/Manual/UnderstandingAutomaticMemoryManagement.html

General Memory Management

• Avoid memory allocation in a loops

function ConcatExample(intArray: int[]) {

var line = intArray[0].ToString();

for (i = 1; i < intArray.Length; i++) {

line += ", " + intArray[i].ToString();

}

return line;

}

General Memory Management

• Avoid memory allocation in a loops

var scoreBoard: GUIText;var score: int;

function Update() {var scoreText: String = "Score: " + score.ToString();scoreBoard.text = scoreText;

}

var scoreBoard: GUIText;var scoreText: String;var score: int;var oldScore: int;

function Update() {if (score != oldScore) {

scoreText = "Score: " + score.ToString();scoreBoard.text = scoreText;oldScore = score;

}}

General Memory Management

• Avoid memory allocation in a loops– Assign random variables to an array.

function RandomList(numElements: int) {

var result = new float[numElements];

for (i = 0; i < numElements; i++) {

result[i] = Random.value;

}

return result;

}

function RandomList(arrayToFill: float[]) {

for (i = 0; i < arrayToFill.Length; i++) {

arrayToFill[i] = Random.value;

}

}

General Memory Management

• Free Memory in C++ and C# (for example: arrays, images, complex objects, ..etc.)

– C++: delete keyword

– C#: enforce garbage collection cleanup