Simulation Engines TDA571|DIT030 3D Graphics - Part 1 Tommaso - - PowerPoint PPT Presentation

Simulation Engines TDA571|DIT030 3D Graphics - Part 1

Tommaso Piazza

1

IDC | Interaction Design Collegium

3D Graphics

2

IDC | Interaction Design Collegium

Once again...

• Real-time 3D graphics is all about triangles
• Properties of a renderable triangle
• Vertices
• Normals
• Vertex order
• (Tangents)
• (Binormals/Bitangents)
• (...)
• 3D graphics is the art of coloring triangles.

3

IDC | Interaction Design Collegium

3D Graphics is also about...

• Vectors
• Matrices
• Quaternions
• Loads and loads of mathematical operations
• n the above.
• If you feel that you are shaky on vector

algebra, take a look in your old math books. You will thank yourself later.

4

IDC | Interaction Design Collegium

Categorization of 3D entities

• Static
• Objects that are immutable
• Good candidates for extreme optimization
• Dynamic
• Moves or changes over time
• Harder to optimize
• Animated
• Could be either static or dynamic
• The vertex object can not be locked

5

IDC | Interaction Design Collegium

Spatial data structures

• Demands on real-time graphics grow rapidly
• Modern graphics hardware is:
• Fast
• Programmable
• Has plenty of memory
• However...
• ”The fastest polygon to render is the one never

sent down to the accelerator's pipeline.”

6

IDC | Interaction Design Collegium

Words of wisdom

• Performances are
• "25k batches/s @ 100% 1GHz CPU"
• To run smoothly
• (i 60fps, 417 batches)

7

IDC | Interaction Design Collegium

Spatial data structures

• A spatial data structure:
• Orders geometric data by its location
• Is often hierarchic, which allows O(log n) searches

and rapidly discarding irrelevant data

• Spatial data structures can be expensive to
• construct. Construction is often preprocessed
• Separate structures for static and dynamic data

8

IDC | Interaction Design Collegium

Spatial Data Structures

• Sample applications
• Collision detection
• Location queries
• Chemical simulations
• Rendering

9

IDC | Interaction Design Collegium

Bounding volume hierarchies

• In general: better fit = more memory and

more expensive intersection algorithm

• Axis-Aligned Bounding Boxes (AABB)
• The simplest type of bounding box
• Enclosed by a model's minimum and maximum

values along the x-, y- and z-axis

• Ordinarily described by a center and the distance

from the center to the edge along each axis

10

IDC | Interaction Design Collegium

Bounding volume hierarchies

• Oriented Bounding Boxes (OBB)
• Similar to AABB's, but rotated to fit the model more

tightly

• Ordinarily described by a center, three normalized

vectors and the distance from the center to the edges

• k-Discrete Oriented Polytope (k-DOP)
• Consists of k/2 normalized plane-normals associated

with two scalars. Together they describe the volume between the two planes

• The k-DOP is the volume enclosed by all planes

11

IDC | Interaction Design Collegium

Bounding volume hierarchies

http://udn.epicgames.com/Two/CollisionTutorial.html

12

IDC | Interaction Design Collegium

Outdoor scenery

• Outdoor scenery is rarely represented in the

same fashion as indoor scenery

• The main problem is the scale of outdoor

scenery

• Outdoor scenery often needs special

considerations for optimization

• Continuous level of detail

13

IDC | Interaction Design Collegium

Heightmap

• The most common representation of a terrain
• Generally stored in a grayscale 2D texture
• Often divided into several smaller objects and

rendered as series of triangle strips

14

IDC | Interaction Design Collegium

Cracks in the terrain

• Beware of cracks in the terrain when

rendering with non-uniform tessellation

• Different levels of detail
• Limited numerical accuracy in 3D card or API
• Can be avoided by drawing “skirts”
• Multiple techniques for this...

15

IDC | Interaction Design Collegium

The ROAM algorithm

• The Realtime Optimally Adapting Meshes

algorithm was designed as a high- performance, flexible terrain rendering algorithm with continuous LOD

• Uses a special data structure called a “triangle

bintree” and two priority queues that drive the split and merge operations of patches in the heightmap based on the location of the camera

16

IDC | Interaction Design Collegium

Trends in terrain rendering

• On modern hardware, algorithms like ROAM

use unnecessarily much CPU-power

• Fillrate and triangle-drawing capabilities on

modern 3D hardware are very good

• Nowadays, terrain rendering is mostly done by

grouping blocks of terrain patches into vertex buffer objects of just a few detail levels, and culled using only view frustum culling

17

IDC | Interaction Design Collegium

Volumetric terrain rendering

• Uses voxels (volume pixels)
• Can be used with 3D scans
• Can be used to render heightmaps
• Potentially very high resolution
• Used in Comanche: Maximum overkill, 1992
• No hardware support

18

IDC | Interaction Design Collegium

Skyboxes and skydomes

• Used to render the background of 3D scenes
• Waste of resources to draw small background objects as

meshes

• Drawn in the following way
• Clear depth and color buffers
• Apply camera transformation
• Disable the depth buffer test

and writes

• Draw the skybox or skydome
• Enable depth buffer test and

writes

• Draw the rest of the scene

19

IDC | Interaction Design Collegium

Indoor scenery

• Data is organized in a very different manner

than for outdoor scenery

• Very different scales than for outdoor scenery
• The existence of enclosing walls and ceilings

give excellent opportunities for culling

• Thanks to the smaller scale, requirements on

detail are much greater than outdoors

• Tim Sweeney estimates that we have a factor
• f 10.000 to 40.000 to go before we are able to

render photo-realistic indoor environments

20

IDC | Interaction Design Collegium

Uniform grids

• The world is split into uniformly

sized voxels (in 3D) or squares (in 2D)

• Very easy to implement
• Relatively efficient with ray-grid-

intersection tests, therefore often used for raytracing

• http://www.cs.yorku.ca/~amana/

research/grid.pdf

• Useful for dynamic data

21

IDC | Interaction Design Collegium

Binary space partitioning trees

• Have been popular since they were used in

DOOM

• Was originally designed to solve depth sorting

in the late 60's.

• Sorts geometry by splitting along planes
• Axis-aligned
• Also known as KD-tree.
• Ordinarily built by alternating the splitting axis.
• Polygon-aligned
• The most commonly used variant in games

22

IDC | Interaction Design Collegium

Quad trees

• Two-dimensional

representation

• Similar to KD-trees, but

always split in the middle

• Useful for representing
• utdoor environments which

are represented with height- maps

23

IDC | Interaction Design Collegium

Octtrees

• Like quad trees, but in 3D
• Efficient for large three-

dimensional spaces

• Often have criteria for how

many children are needed in a voxel before splitting it as well as how many levels of hierarchy are allowed

• Structures where the voxels

do not necessarily have the same size are called loose

• ctrees/quadtrees

24

IDC | Interaction Design Collegium

Culling

• Culling is the act of removing polygons that

do not need to be rendered

• Can be done by either the CPU or the GPU
• Culling on CPU generally saves bandwidth
• Culling on GPU (probably) saves CPU-time
• Modern hardware stores geometry in vertex

buffers, and encourages culling per object instead of per polygon, because that allows reusing of vertex buffer objects

25

IDC | Interaction Design Collegium

Software vs. Hardware culling

GPU Viewport clipping Z-buffer testing Backface culling CPU View frustum culling Occlusion culling Backface culling

26

IDC | Interaction Design Collegium

Backface culling

• Ordinarily done in the

geometry stage of the

• pipeline. All data is sent to

the GPU, but only visible pixels are rasterized

• A polygon is backfacing if

n·v < 0

• The amount of rendered

triangles is normally reduced by 50%

27

IDC | Interaction Design Collegium

View frustum culling

• Culls objects that are
• utside of the view frustum
• The view frustum is volume

enclosed by six planes that is visible from the camera

• Very efficient when used on a

hierarchy of spatial data structure

28

IDC | Interaction Design Collegium

Example: View frustum culling over an octtree

• The view frustum is represented by six planes

that point to the inside of the enclosed volume

• Recurse over the following algorithm for voxels

in the octtree, beginning with the voxel that contains the entire tree:

• If the voxel is outside of any of the planes, do

not draw it

• If the voxel is inside all of the planes, draw all of

its contents

• If the voxel intersects any of the planes, perform

this test for all of its children

29

IDC | Interaction Design Collegium

Occlusion culling

• Performed per pixel by the Z-buffer by the GPU
• View frustum culling does not discard
• bjects which are hidden behind other
• bjects
• With frustum culling and similar methods, data

structures overestimate the size of objects. With

• cclusion culling, silhouettes which

underestimate sizes are used

• There are hardware extensions for occlusion

culling

30

IDC | Interaction Design Collegium

Portal culling

• A special case of occlusion culling.
• Data is partitioned into rooms. Their

insides are only visible through portals (windows, doors, etc.).

• Frustum culling is recursively

performed on the volumes that are created between the camera and the portal

• Ordinarily, engines use either portal- or

BSP-structures

• Artist controlled

31

IDC | Interaction Design Collegium

Antiportals

• Relatively modern technique.
• Artist-controlled method of occlusion culling
• Antiportals are similar to portals, but instead of

revealing objects, data within the frustum between the camera and the antiportal is culled.

• Used by the Unreal Engine from UT2003 and
• nward.

32

IDC | Interaction Design Collegium

Level of detail

• LOD is based on the assumption that fewer

triangles are needed to visualize an object when it is further away from the camera

• Two alternative ways:
• Static: several precreated alternative models are

used.

• Dynamic: a single model is altered in runtime.
• LOD is also applicable for materials.

33

IDC | Interaction Design Collegium

Static LOD

• One of several pre-created models is drawn

depending on the distance to it

• Fast and simple but requires more memory and

more work with creating alternative models

• Artifacts may arise when switching models. This

can be solved by blending between the models (preferably in a vertex program on the GPU)

• Be careful not to be counterproductive. Too

many levels or distance functions may result in even more data being sent to the GPU

34

IDC | Interaction Design Collegium

Dynamic LOD

• Less ”popping” than with LOD, but

likely creates models that look worse than those created by artists

• Uses more CPU-time than static LOD
• Two popular methods:
• Continuous Level of Detail (CLOD):

Removes the edge between two triangles when its two vertices are close to each

• ther
• Geomorph Level of Detail: Builds on

CLOD, but interpolates intermediate positions for corners before merging

35

IDC | Interaction Design Collegium

Static objects

• Static objects in 3D have immutable

geometry and are only subject to affine transformations

• Possible optimizations
• Storage
• Since the geometry never changes, we only need one

copy of each object, even if we have multiple instances of it in our scene

• Rendering
• The static nature of the geometry allows for rendering
• ptimizations such as display lists and vertex arrays and

buffers

36

IDC | Interaction Design Collegium

Vertex arrays/objects/buffers

• The traditional OpenGL way of rendering consist
• f a lot of glBegin, glEnd, glVertex, glNormal, etc.
• Unnecessary work
• OpenGL introduced vertex arrays
• Vertex buffers in Direct3D
• Specifies a block of vertices for reuse
• Thanks to hardware T&L (transform and lightning) and

locking the vertex array, the 3D card can perform most of this work without CPU intervention

• OpenGL uses VBOs nowadays

37

IDC | Interaction Design Collegium

Characters

• Characters tend to be subject to vertex

animation

• Vertex arrays can not be locked
• Often animated with use of skeletal animation
• Can be done almost entirely in hardware, which

allows for locked arrays

• Morph targets
• Harder to do in hardware, but optimizations can be

done

38

IDC | Interaction Design Collegium

Data structures in Ogre

• Ogre allows you to choose between multiple

Scene Managers which use different data structures.

• Octtree
• Terrain: Uses an octtree for objects and a grid for the

terrain.

• Nature Scene Manager (ogreaddons)
• Paging Scene Manager (ogreaddons)
• BSP Scene Manager: Certain legal problems.
• Dot Scene Manager (ogreaddons): Recommended

instead of BSP in the documentation. Uses an octtree.

39

IDC | Interaction Design Collegium

Culling in Ogre

• A few quotes from the forum:
• ”The scene manager culling method. This is customized based
• n the scene manager being used; for example the BSP

manager culls based on bsp leaves, the octree manager culls based on an octree etc. Normally these managers cull at a high level though, bounding boxes typically.”

• ”Frustum culling. Once the scene manager has identified

potentially visible chunks, a basic frustum cull is done based on the bounding box.”

• ”GPU culling. The GPU will eliminate backface triangles (unless

you change the culling mode in the .material) and offscreen triangles will be discarded during primitive assembly.”

40

IDC | Interaction Design Collegium

Ogre tips

• Read about the Ogre scene managers at:

http://www.ogre3d.org/wiki/index.php/ SceneManagersFAQ

• Read about terrains in Ogre at:

http://www.ogre3d.org/wiki/index.php/ TerrainHowto

• Siggraph papers can be read for free from

Chalmers computers at: http://portal.acm.org/portal.cfm

• Google-equivalent:

http://scholar.google.se/

41