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

Simulation Engines TDA571|DIT030 3D Graphics - Part 1 Tommaso Piazza 1

3D Graphics IDC | Interaction Design Collegium 2

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. IDC | Interaction Design Collegium 3

3D Graphics is also about...  Vectors  Matrices  Quaternions  Loads and loads of mathematical operations on 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. IDC | Interaction Design Collegium 4

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 IDC | Interaction Design Collegium 5

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 .” IDC | Interaction Design Collegium 6

Words of wisdom  Performances are  "25k batches/s @ 100% 1GHz CPU"  To run smoothly  (i 60fps, 417 batches) IDC | Interaction Design Collegium 7

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 IDC | Interaction Design Collegium 8

Spatial Data Structures  Sample applications  Collision detection  Location queries  Chemical simulations  Rendering IDC | Interaction Design Collegium 9

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 IDC | Interaction Design Collegium 10

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 IDC | Interaction Design Collegium 11

Bounding volume hierarchies http://udn.epicgames.com/Two/CollisionTutorial.html IDC | Interaction Design Collegium 12

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 IDC | Interaction Design Collegium 13

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 IDC | Interaction Design Collegium 14

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... IDC | Interaction Design Collegium 15

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 IDC | Interaction Design Collegium 16

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 IDC | Interaction Design Collegium 17

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 IDC | Interaction Design Collegium 18

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 IDC | Interaction Design Collegium 19

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 of 10.000 to 40.000 to go before we are able to render photo-realistic indoor environments IDC | Interaction Design Collegium 20

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 IDC | Interaction Design Collegium 21

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 IDC | Interaction Design Collegium 22

Quad trees  Two-dimensional representation  Similar to KD-trees, but always split in the middle  Useful for representing outdoor environments which are represented with height- maps IDC | Interaction Design Collegium 23

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 octrees/quadtrees IDC | Interaction Design Collegium 24

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 IDC | Interaction Design Collegium 25

Software vs. Hardware culling GPU CPU Viewport clipping View frustum culling Z-buffer testing Occlusion culling Backface culling Backface culling IDC | Interaction Design Collegium 26

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% IDC | Interaction Design Collegium 27

View frustum culling  Culls objects that are outside 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 IDC | Interaction Design Collegium 28

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 IDC | Interaction Design Collegium 29

Occlusion culling  Performed per pixel by the Z-buffer by the GPU  View frustum culling does not discard objects which are hidden behind other objects  With frustum culling and similar methods, data structures overestimate the size of objects. With occlusion culling, silhouettes which underestimate sizes are used  There are hardware extensions for occlusion culling IDC | Interaction Design Collegium 30