OpenGL Performer Programmer’s Guide Version 3.2 007-1680-100

OpenGL Performer Programmer’s Guide Version 3.2 007-1680-100
OpenGL Performer™
Programmer’s Guide
Version 3.2
007-1680-100
CONTRIBUTORS
Written by George Eckel and Ken Jones
Illustrated by Chrystie Danzer and Chris Wengelski
Production by Karen Jacobson
Engineering contributions by Angus Dorbie, Paolo Farinelli, Tom Flynn, Yair Kurzion, Radomir Mech, Alexandre Naaman, Marcin
Romaszewicz, Stace Peterson, Allan Schaffer, and Jenny Zhao
COPYRIGHT
© 1994, 2000–2004 Silicon Graphics, Inc. All rights reserved; provided portions may be copyright in third parties, as indicated elsewhere herein.
No permission is granted to copy, distribute, or create derivative works from the contents of this electronic documentation in any manner, in
whole or in part, without the prior written permission of Silicon Graphics, Inc.
LIMITED RIGHTS LEGEND
The software described in this document is "commercial computer software" provided with restricted rights (except as to included open/free
source) as specified in the FAR 52.227-19 and/or the DFAR 227.7202, or successive sections. Use beyond license provisions is a violation of
worldwide intellectual property laws, treaties and conventions. This document is provided with limited rights as defined in 52.227-14.
TRADEMARKS AND ATTRIBUTIONS
Silicon Graphics, SGI, the SGI logo, IRIS, IRIX, ImageVision Library, Indigo, Indy, InfiniteReality, O2, Octane, Onyx, Onyx2, and OpenGL are
registered trademarks and CASEVision, Crimson, Elan Graphics, IRIS Geometry Pipeline, IRIS GL, IRIS Graphics Library, IRIS InSight,
IRIS Inventor, Indigo Elan, Indigo2, InfinitePerformance, InfiniteReality2, InfiniteReality4, Onyx4, OpenGL Multipipe, OpenGL Performer,
Performance Co-Pilot, REACT, RealityEngine, RealityEngine2, Showcase, Silicon Graphics Prism, UltimateVision, and VPro are trademarks of
Silicon Graphics, Inc., in the United States and/or other countries worldwide.
AutoCAD is a registered trademark of Autodesk, Inc. CATIA is a registered trademark of DASSAULT SYSTEMES S.A. Designer’s Workbench is
a trademark of Centric Software, Inc. Lightscape is a trademark of Autodesk, Inc. Linux is a registered trademark of Linus Torvalds. Maya is a
registered trademark and Wavefront is a trademark of Alias Systems, a division of Silicon Graphics Limited in the United States and/or other
countries worldwide. Motif is a registered trademark and X Window System and OSF/Motif are trademarks of The Open Group. Purify is a
registered trademark of Rational Software Corporation. Red Hat is a registered trademark of Red Hat, Inc. RPC is a trademark of ArchVision.
VTune is a trademark of Intel Corporation. WindView is a trademark of Wind River Systems. Microsoft, Windows, and Windows NT are
registered trademarks of Microsoft Corporation in the United States and other countries. Maya is a trademark of Alias Systems. All other
trademarks mentioned herein are the property of their respective owners.
PATENT DISCLOSURE
Many of the techniques and methods disclosed in this Programmer’s Guide are covered by patents held by Silicon Graphics including U.S.
Patent Nos. 5,051,737; 5,369,739; 5,438,654; 5,394,170; 5,528,737; 5,528,738; 5,581,680; 5,471,572 and patent applications pending.
We encourage you to use these features in your OpenGL Performer application on SGI systems.
This functionality and OpenGL Performer are not available for re-implementation and distribution on other platforms without the explicit
permission of Silicon Graphics.
New Features in This Guide
This revision of the guide documents OpenGL Performer 3.2, which has support for the
following features:
007-1680-100
•
Silicon Graphics Prism visualization systems
•
OpenGL 2.0 Shading Language
•
Scene graph optimizer
•
New compositor API
•
Maya 6.0 exporter
iii
Record of Revision
Version
Description
020
1994
Original publication.
007-1680-100
060
November 2000
Updated for the 2.4 version of OpenGL Performer.
070
November 2001
Updated for the 2.5 version of OpenGL Performer.
080
December 2002
Updated for the 3.0 version of OpenGL Performer.
090
December 2003
Updated for the 3.1 version of OpenGL Performer.
100
November 2004
Updated for the 3.2 version of OpenGL Performer.
v
Contents
Figures .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Tables .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. xxxvii
Examples
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . xlv
. . xlv
. xlvi
. xlvi
. xlvi
. xlviii
. xlix
. xlix
. . l
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
About This Guide. . . . . . . . . . . . . .
Why Use OpenGL Performer? . . . . . . . . . .
What You Should Know Before Reading This Guide . . . .
How to Use This Guide . . . . . . . . . . . .
What This Guide Contains . . . . . . . . . .
Sample Applications . . . . . . . . . . . .
Conventions . . . . . . . . . . . . . .
Internet and Hardcopy Reading for the OpenGL Performer Series
Reader Comments . . . . . . . . . . . . . .
1.
007-1680-100
OpenGL Performer Programming Interface
General Naming Conventions . . . .
Prefixes. . . . . . . . . .
Header Files . . . . . . . .
Naming in C and C++ . . . . .
Abbreviations . . . . . . . .
Macros, Tokens, and Enums. . . .
Class API . . . . . . . . . .
Object Creation . . . . . . .
Set Routines . . . . . . . .
Get Routines . . . . . . . .
Action Routines . . . . . . .
Enable and Disable of Modes . . .
Mode, Attribute, or Value . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
xxxi
. xli
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
1
2
2
3
3
3
4
4
4
5
5
6
vii
Contents
Base Classes . . . . . . . .
Inheritance Graph . . . . .
libpr and libpf Objects . .
User Data . . . . . . .
pfDelete() and Reference Counting
Copying Objects with pfCopy() .
Printing Objects with pfPrint() .
Determining Object Type . . .
viii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2.
Setting Up the Display Environment . . . . . . .
Using Pipes . . . . . . . . . . . . . . .
The Functional Stages of a Pipeline . . . . . . .
Creating and Configuring a pfPipe . . . . . . .
Example of pfPipe Use . . . . . . . . . .
Using Channels . . . . . . . . . . . . . .
Creating and Configuring a pfChannel . . . . . .
Setting Up a Scene . . . . . . . . . . . .
Setting Up a Viewport . . . . . . . . . .
Setting Up a Viewing Frustum . . . . . . . .
Setting Up a Viewpoint . . . . . . . . . .
Example of Channel Use . . . . . . . . . .
Controlling the Video Output. . . . . . . . . .
Using Multiple Channels . . . . . . . . . . .
One Window per Pipe, Multiple Channels per Window .
Using Channel Groups. . . . . . . . . . . .
Multiple Channels and Multiple Windows . . . . .
Importing OpenGL Multipipe SDK (MPK) Configuration Files
3.
Nodes and Node Types
Nodes . . . . .
Attribute Inheritance
pfNode . . . .
pfGroup. . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 6
. 7
. 10
. 10
. 11
. 15
. 15
. 17
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 19
. 21
. 21
. 23
. 25
. 26
. 26
. 27
. 27
. 28
. 30
. 32
. 34
. 35
. 36
. 40
. 43
. 44
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 49
. 49
. 49
. 51
. 53
007-1680-100
Contents
Working with Nodes . .
Instancing . . . .
Bounding Volumes .
Node Types. . . . .
pfScene Nodes. . .
pfSCS Nodes . . .
pfDCS Nodes . . .
pfFCS Nodes . . .
pfDoubleSCS Nodes .
pfDoubleDCS Nodes .
pfDoubleFCS Nodes .
pfSwitch Nodes . .
pfSequence Nodes .
pfLOD Nodes . . .
pfASD Nodes . . .
pfLayer Nodes . .
pfGeode Nodes . .
pfText Nodes . . .
pfBillboard Nodes .
pfPartition Nodes . .
Sample Program . . .
4.
007-1680-100
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
56
56
59
61
61
62
62
63
64
64
67
67
67
70
70
70
71
73
75
78
80
Database Traversal . . . . . . . .
Scene Graph Hierarchy . . . . . . .
Database Traversals . . . . . . .
State Inheritance . . . . . . . .
Database Organization . . . . . .
Application Traversal . . . . . . . .
Cull Traversal . . . . . . . . . .
Traversal Order . . . . . . . .
Visibility Culling . . . . . . . .
Organizing a Database for Efficient Culling
Sorting the Scene . . . . . . . .
Paths through the Scene Graph . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 85
. 87
. 87
. 87
. 88
. 88
. 90
. 90
. 91
. 94
. 97
.100
ix
Contents
Draw Traversal . . . . . . . . .
Optimizing the Drawing of Sub-bins .
Bin Draw Callbacks . . . . . .
Controlling and Customizing Traversals .
pfChannel Traversal Modes . . . .
Cull Programs . . . . . . . .
pfNode Draw Mask . . . . . .
pfNode Cull and Draw Callbacks . .
Process Callbacks . . . . . . . .
Process Callbacks and Passthrough Data
Intersection Traversal . . . . . . .
Testing Line Segment Intersections . .
Intersection Requests: pfSegSets . . .
Intersection Return Data: pfHit Objects .
Intersection Masks. . . . . . .
Discriminator Callbacks . . . . .
Line Segment Clipping . . . . .
Traversing Special Nodes. . . . .
Picking . . . . . . . . . .
Performance . . . . . . . .
Intersection Methods for Segments . .
5.
x
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
100
101
101
102
102
103
112
113
116
118
120
120
121
121
122
124
125
125
126
126
127
Frame and Load Control . . . . . . . .
Frame Rate Management . . . . . . . .
Selecting the Frame Rate . . . . . . .
Achieving the Frame Rate . . . . . .
Fixing the Frame Rate . . . . . . . .
Level-of-Detail Management . . . . . . .
Level-of-Detail Models . . . . . . .
Level-of-Detail States . . . . . . . .
Level-of-Detail Range Processing . . . .
Level-of-Detail Transition Blending . . . .
Run-Time User Control Over LOD Evaluation.
Terrain Level-of-Detail . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
129
129
130
130
131
136
136
140
141
145
146
147
007-1680-100
Contents
6.
007-1680-100
Maintaining Frame Rate Using Dynamic Video Resolution . . .
The Channel in DVR . . . . . . . . . . . . .
DVR Scaling . . . . . . . . . . . . . . .
Customizing DVR . . . . . . . . . . . . .
Understanding the Stress Filter . . . . . . . . . .
Dynamic Load Management . . . . . . . . . . . .
Successful Multiprocessing with OpenGL Performer . . . . .
Review of Rendering Stages . . . . . . . . . . .
Choosing a Multiprocessing Model. . . . . . . . .
Asynchronous Database Processing . . . . . . . .
Placing Multiple OpenGL Performer Processes on a Single CPU
Rules for Invoking Functions While Multiprocessing . . .
Multiprocessing and Memory . . . . . . . . . .
Shared Memory and pfInit() . . . . . . . . . .
pfDataPools . . . . . . . . . . . . . . .
Passthrough Data . . . . . . . . . . . . . .
CULL Process Optimizations. . . . . . . . . . . .
Cull Sidekick Processes . . . . . . . . . . . .
Configuring CULL_SIDEKICK Processes . . . . . . .
CULL Sidekick Optimization Mask. . . . . . . . .
CULL Sidekick Synchronization Policy . . . . . . .
CULL Sidekick User Functions . . . . . . . . . .
Modifying Attributes of Cloned pfGeoSets . . . . . .
Marking pfGeoSets for Optimization . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.148
.149
.150
.150
.151
.152
.155
.155
.156
.163
.165
.166
.168
.169
.170
.170
.171
.173
.174
.174
.175
.176
.178
.179
Creating Visual Effects .
Using pfEarthSky . . .
Atmospheric Effects . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.181
.181
.182
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
xi
Contents
Patchy Fog and Layered Fog . . . . . . . . .
Creating Layered Fog . . . . . . . . . .
Creating Patchy Fog . . . . . . . . . .
Initializing a pfVolFog . . . . . . . . .
Updating the View . . . . . . . . . .
Drawing a Scene with Fog . . . . . . . .
Deleting a pfVolFog . . . . . . . . . .
Specifying Fog Parameters . . . . . . . .
Advanced Features of Layered Fog and Patchy Fog .
Performance Considerations and Limitations . . .
Real-Time Shadows . . . . . . . . . . .
Creating a pfShadow . . . . . . . . . .
Drawing a Scene with Shadows . . . . . . .
Specifying Shadow Parameters . . . . . . .
Assigning Data with Directions . . . . . . .
Limitations of Real-Time Shadows . . . . . .
Image-Based Rendering . . . . . . . . . .
Creating a pfIBRnode . . . . . . . . . .
Creating a pfIBRnode Using a Proxy . . . . .
Creating a pfIBRtexture . . . . . . . . .
Parameters Controlling Drawing of a pfIBRnode . .
The Simplify Application . . . . . . . . .
Creating Images of an Object with makeProxyImages
Creating Images of an Object with makeIBRimages .
Limitations . . . . . . . . . . . . .
7.
xii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Importing Databases . . . . . . . . . . . . . . . . . .
Overview of OpenGL Performer Database Creation and Conversion . . . .
libpfdu - Utilities for Creating Efficient OpenGL Performer Run-Time Structures
pfdLoadFile - Loading Arbitrary Databases into OpenGL Performer . . .
Database Loading Details. . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
186
187
187
188
191
191
191
192
193
196
198
199
201
202
202
203
204
204
205
206
208
209
215
219
220
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
221
221
222
223
225
007-1680-100
Contents
Developing Custom Importers . . . . . . . . . . . . . . . .
Structure and Interpretation of the Database File Format . . . . . . .
Scene Graph Creation Using Nodes as Defined in libpf . . . . . . .
Defining Geometry and Graphics State for libpr . . . . . . . . .
Creating an OpenGL Performer Database Converter using libpfdu . . . .
Maximizing Database Loading and Paging Performance with PFB and PFI Formats .
pfconv. . . . . . . . . . . . . . . . . . . . . .
pficonv . . . . . . . . . . . . . . . . . . . . .
Supported Database Formats. . . . . . . . . . . . . . . . .
007-1680-100
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.228
.229
.229
.230
.230
.240
.240
.241
.241
xiii
Contents
Description of Supported Formats . . . . .
AutoDesk 3DS Format . . . . . . .
SGI BIN Format . . . . . . . . .
Side Effects POLY Format . . . . . .
Brigham Young University BYU Format . .
Optimizer CSB Format . . . . . . .
Virtual Cliptexture CT Loader . . . . .
Designer’s Workbench DWB Format . . .
AutoCAD DXF Format . . . . . . .
MultiGen OpenFlight Format . . . . .
McDonnell-Douglas GDS Format . . . .
SGI GFO Format . . . . . . . . .
SGI IM Format . . . . . . . . . .
AAI/Graphicon IRTP Format . . . . .
SGI Open Inventor Format . . . . . .
Lightscape Technologies LSA and LSB Formats
Medit Productions MEDIT Format . . . .
NFF Neutral File Format . . . . . . .
Wavefront Technology OBJ Format . . . .
SGI PFB Format . . . . . . . . .
SGI PFI Format. . . . . . . . . .
SGI PHD Format . . . . . . . . .
SGI PTU Format . . . . . . . . .
ArchVision RPC Format . . . . . . .
USNA Standard Graphics Format . . . .
SGI SGO Format . . . . . . . . .
USNA Simple Polygon File Format . . . .
Sierpinski Sponge Loader. . . . . . .
Star Chart Format . . . . . . . . .
3D Lithography STL Format . . . . . .
SuperViewer SV Format . . . . . . .
Geometry Center Triangle Format . . . .
UNC Walkthrough Format . . . . . .
xiv
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
244
244
244
246
248
249
249
249
250
252
254
254
256
257
258
259
263
263
265
267
268
268
270
273
275
276
280
280
281
282
283
287
287
007-1680-100
Contents
8.
007-1680-100
WRL Format . . . . . . . . . . . . . . .
Database Operators with Pseudo Loaders . . . . . . . .
The Maya Database Exporter. . . . . . . . . . . .
Installation Requirements . . . . . . . . . . .
Exporting a Scene Using the Graphical Interface . . . . .
Exporting a Scene Using the Maya Embedded Language (MEL)
Translation Details . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.287
.288
.290
.290
.291
.296
.297
Geometry . . . . . . . . . .
pfGeoSet (Geometry Set) . . . . . .
Primitive Types . . . . . . .
pfGeoSet Draw Mode . . . . .
Primitive Connectivity . . . . .
Attributes . . . . . . . . .
Attribute Bindings . . . . . .
Indexed Arrays . . . . . . .
pfGeoSet Operations . . . . . .
pfGeoArray (Geometry Array) . . . .
Creating pfGeoArrays . . . . .
pfGeoArray Attributes . . . . .
pfGeoArray Attribute Types. . . .
pfGeoArray Primitive Types . . .
Example Code . . . . . . . .
Converting pfGeoSets to pfGeoArrays
Optimizing Geometry for Rendering . .
Function pfdMergeGraph() . . . .
Function pfdStripGraph() . . . .
Function pfdSpatializeGraph() . . .
The Optimization Pipeline . . . .
Using the libpfgopt Pseudo Loader. .
Rendering 3D Text. . . . . . . .
pfFont . . . . . . . . . .
pfString . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.305
.305
.307
.308
.310
.313
.315
.316
.318
.319
.320
.320
.323
.323
.324
.328
.329
.329
.330
.331
.332
.333
.335
.335
.337
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
xv
Contents
9.
xvi
Higher-Order Geometric Primitives . . . . . . . . . .
Features and Uses of Higher-Order Geometric Primitives . . . .
Reps and the Rendering Process . . . . . . . . . . .
Trimmed NURBS . . . . . . . . . . . . . . .
Objects Required by Reps . . . . . . . . . . . . . .
New Types Required for Reps . . . . . . . . . . .
Classes for Scalar Functions . . . . . . . . . . . .
Matrix Class: pfRMatrix . . . . . . . . . . . . .
Geometric Primitives: The Base Class pfRep and the Application repTest
Class Declaration for pfRep . . . . . . . . . . . .
Main Features of the Methods in pfRep. . . . . . . . .
Planar Curves . . . . . . . . . . . . . . . . .
Mathematical Description of a Planar Curve . . . . . . .
Lines in the Plane . . . . . . . . . . . . . . .
Circles in the Plane . . . . . . . . . . . . . .
Superquadric Curves: pfSuperQuadCurve2d . . . . . . .
Hermite-Spline Curves in the Plane . . . . . . . . . .
NURBS Overview . . . . . . . . . . . . . . .
NURBS Curves in the Plane . . . . . . . . . . . .
Piecewise Polynomial Curves: pfPieceWisePolyCurve2d. . . .
Discrete Curves in the Plane . . . . . . . . . . . .
Spatial Curves . . . . . . . . . . . . . . . . .
Lines in Space . . . . . . . . . . . . . . . .
Circles in Space . . . . . . . . . . . . . . .
Superquadrics in Space . . . . . . . . . . . . .
Hermite Spline Curves in Space . . . . . . . . . . .
NURBS Curves in Space . . . . . . . . . . . . .
Curves on Surfaces: pfCompositeCurve3d . . . . . . . .
Discrete Curves in Space . . . . . . . . . . . . .
Example of Using pfDisCurve3d and pfHsplineCurve3d . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
341
342
342
342
342
343
343
345
346
348
348
348
349
353
354
356
359
360
365
367
368
371
371
372
372
373
373
374
375
375
007-1680-100
Contents
Parametric Surfaces . . . . . . . . . . . . . .
Mathematical Description of a Parametric Surface . . . .
Defining Edges of a Parametric Surface: Trim Loops and Curves
Adjacency Information: pfEdge . . . . . . . . . .
Base Class for Parametric Surfaces: pfParaSurface . . . .
pfPlaneSurface . . . . . . . . . . . . . .
pfSphereSurface . . . . . . . . . . . . . .
pfCylinderSurface . . . . . . . . . . . . .
pfTorusSurface . . . . . . . . . . . . . .
pfConeSurface. . . . . . . . . . . . . . .
Swept Surfaces . . . . . . . . . . . . . .
pfFrenetSweptSurface . . . . . . . . . . . .
Ruled Surfaces . . . . . . . . . . . . . .
Coons Patches . . . . . . . . . . . . . . .
NURBS Surfaces . . . . . . . . . . . . . .
Hermite-Spline Surfaces . . . . . . . . . . . .
Meshes . . . . . . . . . . . . . . . . . .
Mesh Faces. . . . . . . . . . . . . . . .
Mesh Vertices . . . . . . . . . . . . . . .
Subdivision Surfaces . . . . . . . . . . . . . .
Creating a Subdivision Surface . . . . . . . . . .
Loop and Catmull-Clark Subdivisions . . . . . . . .
Dynamic Modification of Vertices . . . . . . . . .
The libpfsubdiv Pseudo Loader . . . . . . . . . .
Special Notes . . . . . . . . . . . . . . .
10.
007-1680-100
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Creating and Maintaining Surface Topology . . . . . . . .
Overview of Topology Tasks . . . . . . . . . . . . .
Summary of Scene Graph Topology: pfTopo . . . . . . . .
Building Topology: Computing and Using Connectivity Information
Reading and Writing Topology Information: Using Pseudo Loaders
Class Declaration for pfTopo . . . . . . . . . . .
Main Features of the Methods in pfTopo . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.375
.376
.377
.379
.380
.384
.386
.389
.391
.393
.395
.399
.400
.402
.404
.411
.413
.416
.417
.420
.421
.424
.424
.424
.425
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.427
.427
.428
.431
.434
.435
.436
xvii
Contents
Collecting Connected Surfaces: pfSolid . .
Class Declaration for pfSolid . . . .
Main Features of the Methods in pfSolid
xviii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
437
437
438
11.
Rendering Higher-Order Primitives: Tessellators . . . .
Features of Tessellators . . . . . . . . . . . .
Tessellators for Varying Levels of Detail . . . . . .
Tessellators Act on a Whole Graph or Single Node . . .
Tessellators and Topology: Managing Cracks . . . . .
Base Class pfTessellateAction . . . . . . . . . . .
Retessellating a Scene Graph . . . . . . . . . .
Class Declaration for pfTessellateAction . . . . . .
Main Features of the Methods in pfTessellateAction . . .
Tessellating Parametric Surfaces . . . . . . . . . .
pfTessParaSurfaceAction . . . . . . . . . . .
Sample From repTest: Tessellating and Rendering a Sphere .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
439
440
441
442
442
443
443
443
444
445
445
448
12.
Graphics State . . . . . . .
Immediate Mode . . . . . .
Rendering Modes . . . . .
Rendering Values . . . . .
Enable / Disable . . . . .
Rendering Attributes . . . .
Graphics Library Matrix Routines
Sprite Transformations . . .
Display Lists . . . . . .
State Management . . . . .
State Override . . . . . .
pfGeoState . . . . . . .
13.
Shaders . . . . . . . . . . .
The pfShaderProgram Class . . . . .
Allocating Memory for a Shader Program
Creating a Shader Program . . . .
Applying Shader Programs . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
451
451
453
458
458
459
477
478
479
480
482
482
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
491
491
492
493
496
007-1680-100
Contents
The pfShaderObject Class. . .
Creating New Shader Objects
Specifying Shader Objects .
Specifying the Object Type .
Compiling Shader Objects .
Example Code . . . . . .
14.
007-1680-100
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Using Scalable Graphics Hardware . . . . . . . . .
Using OpenGL Performer with a DPLEX . . . . . . . .
Hyperpipe Concepts . . . . . . . . . . . . .
Configuring Hyperpipes . . . . . . . . . . . .
Configuring pfPipeWindows and pfChannels . . . . .
Programming with Hyperpipes . . . . . . . . . .
Using OpenGL Performer with an SGI Scalable Graphics Compositor
How the Compositor Functions . . . . . . . . . .
The pfCompositor Class . . . . . . . . . . . .
Querying the System for Hardware Compositors. . . . .
Creating a pfCompositor. . . . . . . . . . . .
Querying pfCompositors . . . . . . . . . . .
Load Balancing . . . . . . . . . . . . . .
Setting Compositor Modes . . . . . . . . . . .
Querying Compositor Modes . . . . . . . . . .
Managing Screen Space, Channel Clipping, and Antialiasing .
Using OpenGL Performer with GPUs . . . . . . . . .
The pfGProgram Class . . . . . . . . . . . .
The pfGProgramParms Class . . . . . . . . . .
The pfVertexProgram and pfFragmentProgram Classes . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.497
.498
.498
.499
.499
.500
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.503
.503
.504
.504
.511
.515
.517
.517
.519
.519
.520
.522
.524
.525
.528
.529
.532
.532
.534
.535
xix
Contents
15.
xx
ClipTextures . . . . . . . . . . . . .
Overview . . . . . . . . . . . . . .
Cliptexture Levels . . . . . . . . . . .
Cliptexture Assumptions . . . . . . . . .
Image Cache . . . . . . . . . . . .
Toroidal Loading . . . . . . . . . . .
Updating the Clipcenter . . . . . . . . .
Virtual Cliptextures on InfiniteReality Systems . .
Cliptexture Support Requirements . . . . . .
Special Features . . . . . . . . . . .
How Cliptextures Interact with the Rest of the System
Cliptexture Support in OpenGL Performer. . . .
Cliptexture Manipulation. . . . . . . . .
Cliptexture API . . . . . . . . . . . . .
Preprocessing ClipTextures . . . . . . . . .
Building a MIPmap . . . . . . . . . .
Formatting Image Data . . . . . . . . .
Tiling an Image . . . . . . . . . . .
Cliptexture Configuration . . . . . . . . . .
Configuration Considerations . . . . . . .
Load-Time Configuration. . . . . . . . .
Post-Load-Time Configuration . . . . . . .
Configuration API . . . . . . . . . . . .
libpr Functionality . . . . . . . . . .
Configuration Utilities . . . . . . . . .
Configuration Files . . . . . . . . . .
Post-Scene Graph Load Configuration . . . . . .
MPClipTextures . . . . . . . . . . .
pfMPClipTexture Utilities . . . . . . . .
Using Cliptextures with Multiple Pipes. . . . .
Texture Memory and Hardware Support Checking .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
537
538
539
540
541
544
546
546
548
548
549
550
552
553
553
554
555
555
556
556
557
557
557
558
561
563
579
579
581
584
586
007-1680-100
Contents
Manipulating Cliptextures . . . . . . .
Cliptexture Load Control . . . . . .
Invalidating Cliptextures . . . . . .
Virtual ClipTextures . . . . . . . .
Custom Read Functions . . . . . . .
Using Cliptextures. . . . . . . . . .
Cliptexture Insets . . . . . . . . .
Estimating Cliptexture Memory Usage. . .
Using Cliptextures in Multipipe Applications.
Virtualizing Cliptextures. . . . . . .
Customizing Load Control . . . . . .
Custom Read Functions . . . . . . .
Cliptexture Sample Code . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.587
.587
.592
.593
.599
.601
.601
.605
.609
.610
.611
.612
.612
16.
Windows . . . . . . . . . . . . . .
pfWindows . . . . . . . . . . . . . .
Creating a pfWindow . . . . . . . . . . .
Configuring the Framebuffer of a pfWindow . . . .
pfWindows and GL Windows . . . . . . . .
Manipulating a pfWindow . . . . . . . . .
Alternate Framebuffer Configuration Windows . .
Window Share Groups . . . . . . . . .
Synchronization of Buffer Swap for Multiple Windows
Communicating with the Window System . . . . .
More pfWindow Examples . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.617
.617
.618
.621
.623
.625
.626
.627
.628
.628
.629
17.
pfPipeWindows and pfPipeVideoChannels . . . .
Using pfPipeWindows . . . . . . . . . .
Creating, Configuring and Opening pfPipeWindow .
pfPipeWindows in Action . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.633
.633
.633
.644
007-1680-100
xxi
Contents
18.
xxii
Controlling Video Displays . . . . . . . .
Creating a pfPipeVideoChannel . . . . . .
Multiple pfPipeVideoChannels in a pfPipeWindow
Configuring a pfPipeVideoChannel . . . . .
Use pfPipeVideoChannels to Control Frame Rate .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
646
647
648
648
649
Managing Nongraphic System Tasks . . .
Handling Queues . . . . . . . . .
Multiprocessing . . . . . . . .
Queue Contents . . . . . . . .
Adding or Retrieving Elements . . . .
pfQueue Modes . . . . . . . .
Running the Sort Process on a Different CPU
High-Resolution Clocks . . . . . . .
Video Refresh Counter (VClock). . . .
Memory Allocation. . . . . . . . .
Allocating Memory With pfMalloc() . .
Shared Arenas . . . . . . . . .
Allocating Locks and Semaphores . . .
Datapools . . . . . . . . . .
CycleBuffers . . . . . . . . .
Asynchronous I/O (IRIX only) . . . . .
Error Handling and Notification . . . . .
File Search Paths . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
651
651
652
652
653
654
657
658
659
659
660
661
662
662
663
666
666
668
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
007-1680-100
Contents
19.
Dynamic Data . . . . . . . . . . .
pfFlux . . . . . . . . . . . . .
Creating and Deleting a pfFlux . . . . .
Initializing the Buffers . . . . . . .
pfFlux Buffers . . . . . . . . . .
Coordinating pfFlux and Connected pfEngines
Synchronized Flux Evaluation . . . . .
Fluxed Geosets . . . . . . . . .
Fluxed Coordinate Systems . . . . . .
Replacing pfCycleBuffer with pfFlux . . .
pfEngine . . . . . . . . . . . .
Creating and Deleting Engines . . . . .
Setting Engine Types and Modes . . . .
Setting Engine Sources and Destinations . .
Setting Engine Masks. . . . . . . .
Setting Engine Iterations . . . . . . .
Setting Engine Ranges . . . . . . .
Evaluating pfEngines. . . . . . . .
Animating a Geometry . . . . . . . .
20.
Active Surface Definition
Overview . . . . .
Using ASD . . . . .
LOD Reduction . .
Hierarchical Structure
ASD Solution Flow Chart .
A Very Simple ASD . .
Morphing Vector . .
A Very Complex ASD
ASD Elements . . . .
Vertices. . . . .
Evaluation Function .
007-1680-100
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.671
.671
.672
.673
.674
.677
.679
.682
.683
.685
.686
.686
.687
.694
.694
.694
.695
.695
.696
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.699
.699
.701
.701
.702
.704
.705
.706
.707
.707
.708
.710
xxiii
Contents
21.
xxiv
Data Structures . . . . . . . . . . . . .
Triangle Data Structure . . . . . . . . .
Attribute Data Array . . . . . . . . . .
Vertex Data Structure . . . . . . . . . .
Default Evaluation Function . . . . . . . .
pfASD Queries . . . . . . . . . . . . .
Aligning an Object to the Surface . . . . . .
Adding a Query Array . . . . . . . . .
Using ASD for Multiple Channels . . . . . . .
Connecting Channels . . . . . . . . . .
Combining pfClipTexture and pfASD . . . . . .
ASD Evaluation Function Timing . . . . . . .
Query Results . . . . . . . . . . . .
Aligning a Geometry With a pfASD Surface Example
Aligning Light Points Above a pfASD Surface Example
Paging . . . . . . . . . . . . . . .
Interest Area . . . . . . . . . . . .
Preprocessing for Paging . . . . . . . . .
Multi-resolution Paging . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
711
713
718
719
720
722
722
723
724
724
725
725
726
726
728
730
730
731
732
Light Points . . . . . . . . .
Uses of Light Points . . . . . .
Creating a Light Point . . . . . .
Setting the Behavior of Light Points . .
Intensity . . . . . . . .
Directionality . . . . . . .
Emanation Shape . . . . . .
Distance. . . . . . . . .
Attenuation through Fog . . . .
Size . . . . . . . . . .
Fading . . . . . . . . .
Callbacks . . . . . . . . .
Multisample, Size, and Alpha . .
Reducing CPU Processing Using Textures
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
735
735
736
737
737
738
738
740
741
742
743
743
746
748
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
007-1680-100
Contents
22.
007-1680-100
Preprocessing Light Points . . . . . . . . . .
Stage Configuration Callbacks . . . . . . . .
How the Light Point Process Works . . . . . .
Calligraphic Light Points . . . . . . . . . . .
Calligraphic Versus Raster Displays . . . . . .
LPB Hardware Configuration . . . . . . . .
Visibility Information . . . . . . . . . .
Required Steps For Using Calligraphic Lights. . . .
Accounting for Projector Differences . . . . . .
Callbacks . . . . . . . . . . . . . .
Frame to Frame Control . . . . . . . . . .
Significance . . . . . . . . . . . . .
Debunching . . . . . . . . . . . . .
Defocussing Calligraphic Objects . . . . . . .
Using pfCalligraphic Without pfChannel . . . . . .
Timing Issues . . . . . . . . . . . . .
Light Point Process and Calligraphic . . . . . .
Debugging Calligraphic Lights on Non-Calligraphic Systems
Calligraphic Light Example . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.749
.749
.750
.751
.752
.754
.756
.757
.759
.761
.763
.764
.764
.765
.765
.766
.766
.766
.767
Math Routines. . . . . . . .
Vector Operations . . . . . . .
Matrix Operations . . . . . . .
Quaternion Operations . . . . .
Matrix Stack Operations . . . . .
Creating and Transforming Volumes .
Defining a Volume . . . . .
Creating Bounding Volumes . .
Transforming Bounding Volumes .
Intersecting Volumes . . . . . .
Point-Volume Intersection Tests .
Volume-Volume Intersection Tests .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.775
.775
.777
.781
.783
.784
.784
.786
.787
.788
.788
.788
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
xxv
Contents
Creating and Working with Line Segments .
Intersecting with Volumes . . . .
Intersecting with Planes and Triangles .
Intersecting with pfGeoSets . . . .
General Math Routine Example Program .
xxvi
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
790
791
792
792
794
23.
Statistics. . . . . . . . . . . . . . . . .
Interpreting Statistics Displays . . . . . . . . . .
Status Line . . . . . . . . . . . . . . .
Stage Timing Graph . . . . . . . . . . . .
Load and Stress . . . . . . . . . . . . .
CPU Statistics . . . . . . . . . . . . . .
Rendering Statistics . . . . . . . . . . . .
Fill Statistics . . . . . . . . . . . . . .
Collecting and Accessing Statistics in Your Application . . .
Displaying Statistics Simply . . . . . . . . . .
Enabling and Disabling Statistics for a Channel . . . .
Statistics in libpr and libpf—pfStats Versus pfFrameStats
Statistics Rules of Use . . . . . . . . . . . .
Reducing the Cost of Statistics . . . . . . . . .
Statistics Output . . . . . . . . . . . . .
Customizing Displays. . . . . . . . . . . .
Setting Update Rate . . . . . . . . . . . .
The pfStats Data Structure . . . . . . . . . .
Setting Statistics Class Enables and Modes . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
799
800
800
801
804
805
807
808
808
809
810
810
811
814
815
817
817
817
818
24.
Performance Tuning and Debugging . . . . .
Performance Tuning Overview . . . . . . .
How OpenGL Performer Helps Performance . . .
Draw Stage and Graphics Pipeline Optimizations .
Cull and Intersection Optimizations. . . . .
Application Optimizations . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
819
819
820
820
822
823
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
007-1680-100
Contents
Specific Guidelines for Optimizing Performance . . . . . .
Graphics Pipeline Tuning Tips . . . . . . . . . .
Process Pipeline Tuning Tips . . . . . . . . . .
Database Concerns . . . . . . . . . . . . .
Special Coding Tips . . . . . . . . . . . . .
Performance Measurement Tools . . . . . . . . . .
Using pixie, prof, and gprof to Measure Performance . .
Using ogldebug to Observe Graphics Calls . . . . . .
Guidelines for Debugging . . . . . . . . . . . .
Shared Memory . . . . . . . . . . . . . .
Use the Simplest Process Model. . . . . . . . . .
Avoid Floating-Point Exceptions . . . . . . . . .
When the Debugger Will Not Give You a Stack Trace . . .
Tracing Members of OpenGL Performer Objects . . . . .
Memory Corruption and Leaks . . . . . . . . . . .
Purify . . . . . . . . . . . . . . . . .
libdmalloc (IRIX only) . . . . . . . . . . .
Notes on Tuning for RealityEngine Graphics . . . . . . .
Multisampling. . . . . . . . . . . . . . .
Transparency . . . . . . . . . . . . . . .
Texturing . . . . . . . . . . . . . . . .
Other Tips . . . . . . . . . . . . . . . .
EventView—A Performance Analyzer . . . . . . . . .
Viewing Events—evanalyzer . . . . . . . . . .
Controlling the Collection of OpenGL Performer Internal Events
Sample Use of EventView . . . . . . . . . . .
Using EventView Tools . . . . . . . . . . . .
Understanding OpenGL Performer Internal Events . . . .
25.
007-1680-100
Building a Visual Simulation Application Using libpfv
Overview . . . . . . . . . . . . . .
The Simplest pfvViewer Program . . . . . . .
Adding Interaction to a pfvViewer Program . . . .
Reading XML Configuration Files . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.824
.824
.828
.832
.836
.837
.837
.838
.839
.839
.840
.841
.841
.842
.842
.843
.844
.844
.845
.845
.846
.846
.847
.847
.849
.849
.853
.860
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.863
.863
.864
.865
.866
xxvii
Contents
26.
xxviii
Module Scoping, Multiple Worlds and Multiple Views . .
Extending a pfvViewer—Writing Custom Modules . . .
Extending a pfvViewer—Module Entry Points . . . . .
Picking, Selection, and Interaction . . . . . . . .
More Sample Programs, Configuration Files, and Source Code
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
870
873
875
876
879
Programming with C++ . . . . . . . . .
Overview . . . . . . . . . . . . .
Class Taxonomy . . . . . . . . . . .
Public Structs . . . . . . . . . . .
libpr Classes. . . . . . . . . . .
libpf Classes . . . . . . . . . . .
pfType Class . . . . . . . . . . .
Programming Basics . . . . . . . . . .
Header Files . . . . . . . . . . .
Creating and Deleting OpenGL Performer Objects
Invoking Methods on OpenGL Performer Objects.
Passing Vectors and Matrices to Other Libraries .
Porting from C API to C++ API . . . . . . .
Typedefed Arrays Versus Structs . . . . .
Interface Between C and C++ API Code . . .
Subclassing pfObjects . . . . . . . . . .
Initialization and Type Definition . . . . .
Defining Virtual Functions . . . . . . .
Accessing Parent Class Data Members . . . .
Multiprocessing and Shared Memory . . . . .
Initializing Shared Memory . . . . . . .
Data Members and Shared Memory. . . . .
Multiprocessing and libpf Objects . . . . .
Performance Hints . . . . . . . . . . .
Constructor Overhead . . . . . . . .
Math Operators . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
881
881
882
882
882
883
883
883
883
887
889
889
889
890
891
891
892
893
894
894
894
896
897
898
898
898
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
007-1680-100
Contents
Glossary
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.899
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.921
Index
007-1680-100
xxix
Figures
Figure 1-1
Figure 2-1
Figure 2-2
Figure 2-3
Figure 2-4
Figure 2-5
Figure 2-6
Figure 2-7
Figure 3-1
Figure 3-2
Figure 3-3
Figure 3-4
Figure 3-5
Figure 4-1
Figure 4-2
Figure 4-3
Figure 4-4
Figure 5-1
Figure 5-2
Figure 5-3
Figure 5-4
Figure 5-5
Figure 5-6
Figure 5-7
Figure 5-8
Figure 6-1
Figure 6-2
007-1680-100
Partial Inheritance Graph of OpenGL Performer Data Types .
From Scene Graph to Visual Display. . . . . . . .
.
.
. 9
. 20
Single Graphics Pipeline . . . . . . . . .
Dual Graphics Pipeline . . . . . . . . .
Symmetric Viewing Frustum . . . . . . . .
Heading, Pitch, and Roll Angles . . . . . . .
Single-Channel and Multiple-Channel Display . . .
The libpfmpk Import Operation . . . . . .
Nodes in the OpenGL Performer Hierarchy . . .
Shared Instances . . . . . . . . . . .
Cloned Instancing . . . . . . . . . . .
A Scenario for Using Double-Precision Nodes . . .
pfDoubleDCS Nodes in a Scene Graph . . . . .
Culling to the Frustum. . . . . . . . . .
Sample Database Objects and Bounding Volumes . .
How to Partition a Database for Maximum Efficiency .
Intersection Methods . . . . . . . . . .
Frame Rate and Phase Control . . . . . . .
Level-of-Detail Node Structure . . . . . . .
Level-of-Detail Processing. . . . . . . . .
Real Size of Viewport Rendered Under Increasing Stress
Stress Processing . . . . . . . . . . .
Multiprocessing Models . . . . . . . . .
Loose Culling of pfGeosets . . . . . . . .
CULL_SIDEKICK Processing . . . . . . . .
Layered Atmosphere Model . . . . . . . .
Patchy Fog Versus Layered Fog . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 22
. 23
. 29
. 31
. 37
. 46
. 50
. 57
. 58
. 65
. 66
. 92
. 94
. 96
.128
.132
.137
.139
.149
.153
.162
.172
.174
.183
.186
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
xxxi
Figures
Figure 6-3
Figure 6-4
Figure 7-1
Figure 7-2
Figure 7-3
Figure 7-4
Figure 7-5
Figure 7-6
Figure 7-7
Figure 7-8
Figure 7-9
Figure 7-10
Figure 7-11
Figure 7-12
Figure 7-13
Figure 7-14
Figure 7-15
Figure 7-16
Figure 8-1
Figure 8-2
Figure 8-3
Figure 8-4
Figure 9-1
Figure 9-2
Figure 9-3
Figure 9-4
Figure 9-5
Figure 9-6
Figure 9-7
Figure 9-8
Figure 9-9
Figure 9-10
xxxii
The Default Simplify Pane . . . . . . . . . . .
The Simplify Pane for Simplifying an Object. . . . . .
BIN-Format Data Objects . . . . . . . . . . .
Soma Cube Puzzle in DWB Form . . . . . . . . .
The Famous Teapot in DXF Form . . . . . . . . .
Spacecraft Model in OpenFlight Format . . . . . . .
GFO Database of Mies van der Rohe’s German Pavilion . .
Aircar Database in IRIS Inventor Format . . . . . . .
LSA-Format City Hall Database . . . . . . . . .
LSB-Format Operating Room Database . . . . . . .
SGI Office Building as OBJ Database . . . . . . . .
Plethora of Polyhedra in PHD Format . . . . . . .
Terrain Database Generated by PTU Tools . . . . . .
Model in SGO Format . . . . . . . . . . . .
Sample STLA Database . . . . . . . . . . . .
Early Automobile in SuperViewer SV Format . . . . .
Maya Export Screen. . . . . . . . . . . . .
Maya Export Options . . . . . . . . . . . .
Primitives and Connectivity . . . . . . . . . .
pfGeoSet Structure . . . . . . . . . . . . .
Indexing Arrays . . . . . . . . . . . . . .
Deciding Whether to Index Attributes . . . . . . .
Class Hierarchy for Higher-Order Primitives . . . . .
Parametric Curve: Parameter Interval (0,1). . . . . . .
Line in the Plane Parameterization . . . . . . . .
Circle in the Plane Parameterization . . . . . . . .
Superquadric Curve’s Dependence on the Parameter α. . .
Hermite Spline Curve Parameterization . . . . . . .
Discrete Curve Definition . . . . . . . . . . .
Parametric Surface: Unit-Square Coordinate System . . .
Trim Loops and Trimmed Surface: Both Trim Loops Made of
Four Trim Curves . . . . . . . . . . . . .
Plane Parameterization . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
210
213
245
250
251
253
255
258
261
262
266
269
271
277
282
284
292
293
312
314
317
318
347
350
353
354
357
359
368
377
.
.
378
384
007-1680-100
Figures
Figure 9-11
Figure 9-12
Figure 9-13
Figure 9-14
Figure 9-15
Figure 9-16
Figure 9-17
Figure 9-18
Figure 9-19
Figure 10-1
Figure 10-2
Figure 11-1
Figure 11-2
Figure 12-1
Figure 12-2
Figure 14-1
Figure 14-2
Figure 14-3
Figure 14-4
Figure 14-5
Figure 14-6
Figure 14-7
Figure 14-8
Figure 14-9
Figure 14-10
Figure 14-11
Figure 15-1
Figure 15-2
Figure 15-3
Figure 15-4
007-1680-100
Sphere Parameterization . . . . . . . . . .
Cylinder Parameterization . . . . . . . . .
Torus Parameterization . . . . . . . . . .
Cone Parameterization. . . . . . . . . . .
Swept Surface: Moving Reference Frame and Effect of
Profile Function. . . . . . . . . . . . .
Ruled Surface Parameterization . . . . . . . .
Coons Patch Construction . . . . . . . . . .
NURBS Surface Control Hull Parameterization
. . .
Hermite Spline Surface With Derivatives Specified at
Knot Points . . . . . . . . . . . . . .
Topological Relations Maintained by Topology Classes .
Consistently Tessellated Adjacent Surfaces and Related
Objects . . . . . . . . . . . . . . .
Class Hierarchy for Tessellators . . . . . . . .
Tessellations Varying With Changes in Control Parameter
pfGeoState Structure . . . . . . . . . . .
Generating the Color of a Multitextured Pixel . . . .
pfPipes Creating pfHyperpipes . . . . . . . .
Multiple Hyperpipes . . . . . . . . . . .
Default Hyperpipe Mapping to Graphic Pipes . . . .
Attaching Objects to the Master pfPipe . . . . . .
Hardware Composition Schemes. . . . . . . .
Horizontal Stripes (pfCompositor Mode) . . . . .
Vertical Stripes (pfCompositor Mode) . . . . . .
Left Tiles (pfCompositor Mode) . . . . . . . .
Right Tiles (pfCompositor Mode) . . . . . . .
Bottom Tiles (pfCompositor Mode) . . . . . . .
Top Tiles (pfCompositor Mode) . . . . . . . .
Cliptexture Components . . . . . . . . . .
Image Cache Components. . . . . . . . . .
Mem Region Update . . . . . . . . . . .
Tex Region Update . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.386
.389
.391
.393
.
.
.
.
.
.
.
.
.396
.401
.403
.407
.
.
.
.
.411
.429
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.430
.440
.441
.487
.488
.505
.506
.507
.512
.518
.526
.527
.527
.527
.528
.528
.538
.539
.542
.543
xxxiii
Figures
Figure 15-5
Figure 15-6
Figure 15-7
Figure 15-8
Figure 15-9
Figure 15-10
Figure 15-11
Figure 15-12
Figure 15-13
Figure 15-14
Figure 17-1
Figure 18-1
Figure 18-2
Figure 19-1
Figure 19-2
Figure 19-3
Figure 19-4
Figure 20-1
Figure 20-2
Figure 20-3
Figure 20-4
Figure 20-5
Figure 20-6
Figure 20-7
Figure 20-8
Figure 20-9
Figure 20-10
Figure 20-11
Figure 20-12
Figure 20-13
Figure 20-14
Figure 20-15
xxxiv
Cliptexture Cache Hierarchy . . . . . . . .
Invalid Border . . . . . . . . . . . .
Clipcenter Moving . . . . . . . . . . .
Virtual Cliptexture Concepts . . . . . . . .
pfMPClipTexture Connections . . . . . . .
pfuClipCenterNode Connections . . . . . . .
Master and Slave Cliptexture Resource Sharing . . .
Cliptexture Insets . . . . . . . . . . .
Supersampled Inset Boundary. . . . . . . .
Offset Slave Tex Regions . . . . . . . . .
Directing Video Output . . . . . . . . .
pfQueue Object . . . . . . . . . . . .
pfCycleBuffer and pfCycleMemory Overview . . .
How pfFlux and Processes Use Frame Numbers . .
pfFlux Buffer Structure . . . . . . . . . .
Timing Diagram Showing the Use of Sync Groups . .
pfEngine Driving a pfFlux That Animates a pfFCS Node
Morphing Range Between LODs . . . . . . .
Large Geometry . . . . . . . . . . . .
ASD Information Flow . . . . . . . . . .
A Very Simple pfASD . . . . . . . . . .
Reference Positions . . . . . . . . . . .
Triangulated Image . . . . . . . . . . .
LOD1 Replaced by LOD2 . . . . . . . . .
Data Structures . . . . . . . . . . . .
ASD Data Structures . . . . . . . . . .
Discontinuous, Neighboring LODs . . . . . .
Triangle Mesh . . . . . . . . . . . .
Using the tsid Field . . . . . . . . . . .
Vertex and Reference Point Arrays, Counter-Clockwise
Ordering . . . . . . . . . . . . . .
Vertex Neighborhoods . . . . . . . . . .
pfASD Evaluation Process . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
544
545
546
547
580
583
584
602
604
610
646
651
665
672
675
681
683
703
704
705
706
709
709
710
711
712
715
715
716
.
.
.
.
.
.
.
.
.
717
720
726
007-1680-100
Figures
Figure 20-16
Figure 20-17
Figure 20-18
Figure 21-1
Figure 21-2
Figure 21-3
Figure 21-4
Figure 21-5
Figure 23-1
Figure 23-2
Figure 23-3
Figure 24-1
Figure 24-2
Figure 24-3
Figure 24-4
Figure 24-5
007-1680-100
Example Setup for Geometry Alignment . . .
Aligning Light Points Above a pfASD Surface . .
Tiles at Different LODs . . . . . . . .
VASI Landing Light . . . . . . . . .
Attenuation Shape . . . . . . . . . .
Attenuation of Light . . . . . . . . .
Lit Multisamples . . . . . . . . . .
Calligraphic Hardware Configuration . . . .
Stage Timing Statistics Display . . . . . .
Conceptual Diagram of a Draw-Stage Timing Line
Other Statistics Classes. . . . . . . . .
The evanalyzer Main Display . . . . . .
User Event myDrawCB . . . . . . . .
Up-Close View of a Single Event . . . . . .
evhist Sample Screen . . . . . . . .
evgraph Sample Screen . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.727
.728
.731
.736
.739
.740
.747
.755
.800
.802
.806
.848
.853
.857
.858
.859
xxxv
Tables
Table 1-1
Table 2-1
Table 3-1
Table 3-2
Table 3-3
Table 3-4
Table 3-5
Table 3-6
Table 3-7
Table 3-8
Table 3-9
Table 3-10
Table 3-11
Table 4-1
Table 4-2
Table 4-3
Table 4-4
Table 4-5
Table 4-6
Table 4-7
Table 4-8
Table 4-9
Table 4-10
Table 5-1
Table 5-2
Table 5-3
Table 5-4
007-1680-100
Routines that Modify libpr Object Reference Counts
Attributes in the Share Mask of a Channel Group . .
OpenGL Performer Node Types . . . . . . .
.
.
.
.
.
.
.
.
.
. 12
. 41
. 51
pfGroup Functions . . . . . . . . . . . . . .
pfDCS Transformations . . . . . . . . . . . .
pfFCS Functions . . . . . . . . . . . . . .
pfSequence Functions . . . . . . . . . . . . .
pfLOD Functions . . . . . . . . . . . . . .
pfLayer Functions . . . . . . . . . . . . . .
pfGeode Functions . . . . . . . . . . . . . .
pfText Functions . . . . . . . . . . . . . .
pfBillboard Functions . . . . . . . . . . . . .
pfPartition Functions . . . . . . . . . . . . .
Traversal Attributes for the Major Traversals . . . . . .
Test Instructions . . . . . . . . . . . . . .
Assign Instructions . . . . . . . . . . . . . .
Jump Instructions . . . . . . . . . . . . . .
Functions Available for User-Defined Cull Program Instructions
Cull Callback Return Values . . . . . . . . . . .
Intersection-Query Token Names . . . . . . . . .
Database Classes and Corresponding Node Masks . . . .
Representing Traversal Mask Values . . . . . . . .
Possible Traversal Results . . . . . . . . . . . .
Frame Control Functions . . . . . . . . . . . .
LOD Transition Zones . . . . . . . . . . . . .
Multiprocessing Models . . . . . . . . . . . .
Trigger Routines and Associated Processes . . . . . . .
. 54
. 63
. 63
. 67
. 70
. 71
. 72
. 73
. 75
. 79
. 86
.106
.106
.107
.109
.113
.121
.123
.124
.125
.131
.145
.156
.168
xxxvii
Tables
Table 6-1
Table 6-2
Table 6-3
Table 6-4
Table 6-5
Table 6-6
Table 6-7
Table 6-8
Table 7-1
Table 7-2
Table 7-3
Table 7-4
Table 7-5
Table 7-6
Table 7-7
Table 7-8
Table 7-9
Table 7-10
Table 7-11
Table 7-12
Table 7-13
Table 7-14
Table 7-15
Table 7-16
Table 7-17
Table 7-18
Table 8-1
Table 8-2
Table 8-3
Table 8-4
Table 8-5
Table 8-6
Table 10-1
xxxviii
pfEarthSky Functions . . . . . . . . . .
pfEarthSky Attributes . . . . . . . . . .
pfVolFog Functions . . . . . . . . . . .
pfVolFog Attributes. . . . . . . . . . .
pfVolFog Flags . . . . . . . . . . . .
pfShadow Functions . . . . . . . . . .
pfShadow Attributes . . . . . . . . . .
Key Command-Line Options of makeProxyImages .
Database-Importer Source Directories . . . . .
libpfdu Database Converter Functions . . . . .
Loader Name Composition . . . . . . . .
libpfdu Database Converter Management Functions.
pfdBuilder Modes and Attributes . . . . . . .
Supported Database Formats . . . . . . . .
Geometric Definitions in LSA Files . . . . . .
RPC Converter Values . . . . . . . . . .
RPC Converter Attributes . . . . . . . . .
Object Tokens in the SGO Format . . . . . . .
Mesh Control Tokens in the SGO Format . . . .
OpenGL Performer Pseudo Loaders . . . . . .
Default Path for the Maya Export Plug-in . . . .
Maya Export Options . . . . . . . . . .
Maya Features Supported by the Exporter . . . .
Maya Exporter Support for UV Mapping Methods . .
Maya Exporter Support for Material Properties . . .
Maya Exporter Support for Texture Properties . . .
pfGeoSet Routines . . . . . . . . . . .
Geometry Primitives . . . . . . . . . .
pfGeoSet PACKED_ATTR Formats . . . . . .
Attribute Bindings . . . . . . . . . . .
pfFont Routines . . . . . . . . . . . .
pfString Routines . . . . . . . . . . .
Topology Building Methods . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
184
184
188
189
190
200
201
215
221
223
225
226
239
242
259
273
275
278
279
289
291
294
297
301
302
303
306
307
310
315
336
339
433
007-1680-100
Tables
Table 10-2
Table 10-3
Table 12-1
Table 12-2
Table 12-3
Table 12-4
Table 12-5
Table 12-6
Table 12-7
Table 12-8
Table 12-9
Table 12-10
Table 12-11
Table 12-12
Table 12-13
Table 13-1
Table 14-1
Table 14-2
Table 14-3
Table 14-4
Table 14-5
Table 14-6
Table 14-7
Table 15-1
Table 15-2
Table 15-3
Table 15-4
Table 15-5
Table 15-6
Table 16-1
Table 16-2
Table 16-3
007-1680-100
Adding Topology and Tessellations to .iv and .csb Files . .
Reading and Writing .pfb Files: with and without Tessellations
pfGeoState Mode Tokens . . . . . . . . . . . .
pfTransparency Tokens . . . . . . . . . . . .
pfGeoState Value Tokens . . . . . . . . . . . .
Enable and Disable Tokens . . . . . . . . . . .
Rendering Attribute Tokens . . . . . . . . . . .
Texture Image Sources . . . . . . . . . . . . .
Texture Load Modes . . . . . . . . . . . . .
Texture Generation Modes . . . . . . . . . . .
pfFog Tokens . . . . . . . . . . . . . . .
pfHlightMode() Tokens . . . . . . . . . . . .
Matrix Manipulation Routines . . . . . . . . . .
pfSprite Rotation Modes . . . . . . . . . . . .
pfGeoState Routines . . . . . . . . . . . . .
Uniform Variable Types . . . . . . . . . . . .
pfPipeWindow Functions That Do Not Propagate . . . . .
Methods for Querying the System for Hardware Compositors .
Methods Used in Creating pfCompositors . . . . . . .
Methods for Querying pfCompositors . . . . . . . .
Static Methods for Querying pfCompositors . . . . . .
Methods to Control the Load Balancing Transitions . . . .
Methods for Managing Screen Space, Channel Clipping, and
Antialiasing . . . . . . . . . . . . . . . .
Tiling Algorithms . . . . . . . . . . . . . .
Image Cache Configuration File Fields . . . . . . . .
Image Tile Filename Tokens . . . . . . . . . . .
Cliptexture Configuration File Fields . . . . . . . .
Parameter Tokens . . . . . . . . . . . . . .
Image Tile Filename Tokens . . . . . . . . . . .
pfWinType() Tokens . . . . . . . . . . . . .
pfWinFBConfigAttrs() Tokens . . . . . . . . . .
Window System Types . . . . . . . . . . . .
.434
.435
.453
.455
.458
.459
.460
.462
.466
.471
.475
.476
.477
.479
.485
.494
.513
.519
.521
.522
.524
.525
.529
.554
.566
.570
.573
.576
.578
.619
.622
.624
xxxix
Tables
Table 16-4
Table 17-1
Table 17-2
Table 18-1
Table 18-2
Table 18-3
Table 18-4
Table 18-5
Table 18-6
Table 18-7
Table 19-1
Table 20-1
Table 21-1
Table 22-1
Table 22-2
Table 22-3
Table 22-4
Table 22-5
Table 22-6
Table 22-7
Table 22-8
Table 22-9
Table 22-10
Table 22-11
Table 22-12
Table 26-1
Table 26-2
Table 26-3
Table 26-4
Table 26-5
Table 26-6
xl
pfWinMode() Tokens . . . . . . . . . . . .
pfPWinType Tokens . . . . . . . . . . . .
Processes From Which to Call Main pfPipeWindow Functions
Thread Information . . . . . . . . . . . . .
Default Input and Output Ranges. . . . . . . . .
pfVClock Routines . . . . . . . . . . . . .
Memory Allocation Routines . . . . . . . . . .
pfNotify Routines . . . . . . . . . . . . .
Error Notification Levels . . . . . . . . . . .
pfFilePath Routines . . . . . . . . . . . . .
pfEngine Types . . . . . . . . . . . . . .
Fields in the Triangle Data Structure . . . . . . . .
Raster Versus Calligraphic Displays . . . . . . . .
Routines for 3-Vectors . . . . . . . . . . . .
Routines for 4x4 Matrices . . . . . . . . . . .
Routines for Quaternions . . . . . . . . . . .
Matrix Stack Routines . . . . . . . . . . . .
Routines to Create Bounding Volumes . . . . . . .
Routines to Extend Bounding Volumes . . . . . . .
Routines to Transform Bounding Volumes . . . . . .
Testing Points for Inclusion in a Bounding Volume. . . .
Testing Volume Intersections . . . . . . . . . .
Intersection Results . . . . . . . . . . . . .
Available Intersection Tests . . . . . . . . . .
Discriminator Return Values . . . . . . . . . .
Corresponding Routines in the C and C++ API . . . . .
Header Files for libpf Scene Graph Node Classes. . . .
Header Files for Other libpf Classes . . . . . . .
Header Files for libpr Graphics Classes . . . . . .
Header Files for Other libpr Classes . . . . . . .
Data and Functions Provided by User Subclasses . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
626
636
641
654
657
659
660
667
667
668
686
713
752
776
777
782
783
786
787
787
788
789
789
793
794
882
883
884
885
886
892
007-1680-100
Examples
Example 1-1
Example 1-2
Example 1-3
Example 1-4
Example 1-5
Example 1-6
Example 2-1
Example 2-2
Example 2-3
Example 2-4
Example 3-1
Example 3-2
Example 3-3
Example 3-4
Example 3-5
Example 3-6
Example 3-7
Example 3-8
Example 3-9
Example 3-10
Example 3-11
Example 4-1
Example 4-2
Example 4-3
Example 4-4
Example 5-1
Example 5-2
007-1680-100
How to Use User Data . . .
Objects and Reference Counts
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Using pfDelete() with libpr Objects . . . . . . . .
Using pfDelete() with libpf Objects . . . . . . . .
Using pfCopy() . . . . . . . . . . . . . . .
General-Purpose Scene Graph Traverser . . . . . . .
pfPipes in Action . . . . . . . . . . . . . .
Using pfChannels . . . . . . . . . . . . . .
Multiple Channels, One Channel per Pipe . . . . . . .
Channel Sharing . . . . . . . . . . . . . .
Making a Scene . . . . . . . . . . . . . . .
Hierarchy Construction Using Group Nodes . . . . . .
Creating Cloned Instances. . . . . . . . . . . .
Automatically Updating a Bounding Volume . . . . . .
Using pfSwitch and pfSequence Nodes . . . . . . . .
Marking a Runway with a pfLayer Node . . . . . . .
Adding pfGeoSets to a pfGeode . . . . . . . . . .
Adding pfStrings to a pfText . . . . . . . . . . .
Setting Up a pfBillboard . . . . . . . . . . . .
Setting Up a pfPartition . . . . . . . . . . . .
Inheritance Demonstration Program . . . . . . . . .
Application Callback to Make a Pendulum . . . . . . .
pfNode Draw Callbacks . . . . . . . . . . . .
Cull-Process Callbacks . . . . . . . . . . . . .
Using Passthrough Data to Communicate with Callback Routines
Frame Control Excerpt . . . . . . . . . . . . .
Setting LOD Ranges . . . . . . . . . . . . .
. 11
. 12
. 13
. 13
. 15
. 17
. 25
. 32
. 39
. 42
. 53
. 55
. 59
. 59
. 68
. 71
. 72
. 73
. 76
. 79
. 80
. 89
.114
.116
.119
.135
.143
xli
Examples
Example 5-3
Example 6-1
Example 6-2
Example 6-3
Example 8-1
Example 8-2
Example 12-1
Example 12-2
Example 12-3
Example 12-4
Example 14-1
Example 14-2
Example 14-3
Example 14-4
Example 14-5
Example 15-1
Example 16-1
Example 16-2
Example 16-3
Example 16-4
Example 17-1
Example 17-2
Example 17-3
Example 17-4
Example 17-5
Example 19-1
Example 19-2
Example 20-1
Example 21-1
Example 21-2
Example 21-3
Example 21-4
xlii
Default Stress Function . . . . . . . . . . . . . 154
How to Configure a pfEarthSky . . . . . . . . . . 182
Fog initialization Using pfVolFogAddPoint() . . . . . . 187
Specifying Patchy Fog Boundaries Using pfVolFogAddNode() . 187
Loading Characters into a pfFont . . . . . . . . . . 336
Setting Up and Drawing a pfString . . . . . . . . . 337
Using pfDecal() to a Draw Road with Stripes . . . . . . 457
Pushing and Popping Graphics State . . . . . . . . . 481
Using pfOverride() . . . . . . . . . . . . . . 482
Inheriting State . . . . . . . . . . . . . . . 484
Configuring a System with Three Hyperpipe Groups . . . . 507
Mapping Hyperpipes to Graphic Pipes . . . . . . . . 508
More Complete Example: Mapping Hyperpipes to Graphic Pipes 508
Set FBConfigAttrs for Each pfPipeWindow . . . . . . . 514
Search the pfPipeWindow List of the pfPipe. . . . . . . 515
Estimating System Memory Requirements . . . . . . . 606
Opening a pfWindow . . . . . . . . . . . . . 618
Using the Default Overlay Window . . . . . . . . . 629
Creating a Custom Overlay Window . . . . . . . . . 630
pfWindows and X Input . . . . . . . . . . . . 631
Creating a pfPipeWindow . . . . . . . . . . . . 634
pfPipeWindow With Alternate Configuration Windows for Statistics
638
Custom Initialization of pfPipeWindow State . . . . . . 640
Configuration of a pfPipeWindow Framebuffer. . . . . . 643
Opening and Closing a pfPipeWindow . . . . . . . . 644
Fluxed pfGeoSet . . . . . . . . . . . . . . . 682
Connecting Engines and Fluxes . . . . . . . . . . 696
Aligning Light Points Above a pfASD Surface . . . . . . 728
Raster Callback Skeleton . . . . . . . . . . . . 745
Preprocessing a Display List - Light Point Process code . . . 750
Setting pfCalligraphic Parameters. . . . . . . . . . 763
Calligraphic Lights . . . . . . . . . . . . . . 767
007-1680-100
Examples
Example 22-1
Example 22-2
Example 22-3
Example 22-4
Example 22-5
Example 26-1
Example 26-2
Example 26-3
Example 26-4
Example 26-5
007-1680-100
Matrix and Vector Math Examples . . . . .
Quaternion Example . . . . . . . . .
Quick Sphere Culling Against a Set of Half-Spaces
Intersecting a Segment With a Convex Polyhedron
Intersection Routines in Action . . . . . .
Valid Creation of Objects in C++ . . . . . .
Invalid Creation of Objects in C++ . . . . .
Class Definition for a Subclass of pfDCS . . .
Overloading the libpf Application Traversal . .
Changeable Static Data Member . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.780
.782
.790
.791
.794
.888
.888
.892
.893
.896
xliii
About This Guide
Welcome to the OpenGL Performer application development environment.
OpenGL Performer provides a programming interface (with ANSI C and C++ bindings)
for creating real-time graphics applications and offers high-performance, multiprocessed
rendering in an easy-to-use 3D graphics toolkit. OpenGL Performer interfaces with the
OpenGL graphics library; this library combined with the IRIX, Linux, or Microsoft
Windows (Windows 2000, Windows NT, and Windows XP) operating system forms the
foundation of a powerful suite of tools and features for creating real-time 3D graphics
applications.
Why Use OpenGL Performer?
Use OpenGL Performer for building visual simulation applications and virtual reality
environments; for rapid rendering in on-air broadcast and virtual-set applications; for
assembly viewing in large, simulation-based design tasks; or to maximize the graphics
performance of any application. Applications that require real-time visuals, free-running
or fixed-frame-rate display, or high-performance rendering will benefit from using
OpenGL Performer.
OpenGL Performer drastically reduces the work required to tune your application’s
performance. General optimizations include the use of highly tuned routines for all
performance-critical operations and the reorganization of graphics data and operations
for faster rendering. OpenGL Performer also handles SGI architecture-specific tuning
issues for you by selecting the best rendering and multiprocessing modes at run time,
based on the system configuration.
OpenGL Performer is an integral part of SGI visual simulation systems. It provides the
interface to advanced features available exclusively with the SGI product line, such as the
Silicon Graphics Prism, Silicon Graphics Onyx4 UltimateVision, InfiniteReality, Silicon
Graphics Octane, Silicon Graphics O2, and VPro graphics subsystems.
OpenGL Performer teamed with SGI graphics hardware provide a sophisticated image
generation system in a powerful, flexible, and extensible software environment. OpenGL
Performer is also tuned to operate efficiently on a variety of graphics platforms; you do
007-1680-100
xlv
About This Guide
not need the hardware sophistication of InfiniteReality graphics to benefit from OpenGL
Performer.
What You Should Know Before Reading This Guide
To use OpenGL Performer, you should be comfortable programming in ANSI C or C++.
You should also have a fairly good grasp of graphics programming concepts. Terms such
as “texture map” and “homogeneous coordinate” are not explained in this guide. It helps
if you are familiar with the OpenGL library.
On the other hand, though you need to know a little about graphics, you do not have to
be a seasoned C (or C++) programmer, a graphics hardware guru, or a graphics-library
virtuoso to use OpenGL Performer. OpenGL Performer puts the engineering expertise
behind SGI hardware and software at your fingertips, so you can minimize your
application development time while maximizing the application’s performance and
visual impact.
For a concise description of OpenGL Performer basics, see the OpenGL Performer Getting
Started Guide.
How to Use This Guide
The best way to get started is to read the OpenGL Performer Getting Started Guide. If you
like learning from sample code, turn to Chapter 1, “Getting Acquainted with OpenGL
Performer,” which takes you on a tour of some demo programs. These programs let you
see for yourself what OpenGL Performer does. Even if you are not developing a visual
simulation application, you might want to look at the demos to see high-performance
rendering in action. At the end of Chapter 2 in that guide, you will find suggestions
pointing to possible next steps; alternatively, you can browse through the summary
below to find a topic of interest.
What This Guide Contains
This guide is divided into the following chapters and appendixes:
•
xlvi
Chapter 1, “OpenGL Performer Programming Interface,” describes the
fundamental ideas behind the OpenGL Performer programming interface.
007-1680-100
About This Guide
007-1680-100
•
Chapter 2, “Setting Up the Display Environment,” describes how to set up
rendering pipelines, windows, and channels (cameras).
•
Chapter 3, “Nodes and Node Types,” describes the data structures used in
OpenGL Performer’s memory-based, scene-definition databases.
•
Chapter 4, “Database Traversal,” explains how to manipulate and examine a scene
graph.
•
Chapter 5, “Frame and Load Control,” explains how to control frame rate,
synchronization, and dynamic load management. This chapter also discusses the
load management techniques of multiprocessing and level-of-detail.
•
Chapter 6, “Creating Visual Effects,” describes how to use environmental,
atmospheric, lighting, and other visual effects to enhance the realism of your
application.
•
Chapter 7, “Importing Databases,” describes database formats and sample
conversion utilities.
•
Chapter 8, “Geometry,” discusses the classes used to create geometry in
OpenGL Performer scenes.
•
Chapter 9, “Higher-Order Geometric Primitives” describes higher-order primitives,
including classes to define discrete curves and surfaces.
•
Chapter 10, “Creating and Maintaining Surface Topology” describes the
connectivity of parametric surfaces—that is, their topology.
•
Chapter 11, “Rendering Higher-Order Primitives: Tessellators” describes how to
control the tessellation of shapes.
•
Chapter 12, “Graphics State,” describes the graphics state, which contains all of the
fields that together define the appearance of geometry.
•
Chapter 13, “Shaders,” describes the shader, a mechanism that allows complex
rendering equations to be applied to OpenGL Performer objects.
•
Chapter 14, “Using Scalable Graphics Hardware,” describes how to use OpenGL
Performer in conjunction with an SGI Video Digital Multiplexer (DPLEX), an SGI
Scalable Graphics Compositor, and graphics processing units (GPUs).
•
Chapter 15, “ClipTextures,” describes how to work with large, high-resolution
textures.
•
Chapter 16, “Windows,” describes how to create, configure, manipulate, and
communicate with a window in OpenGL Performer.
xlvii
About This Guide
•
Chapter 17, “pfPipeWindows and pfPipeVideoChannels,” describes the unified
window and video channel control and management provided by pfPipeWindows
and pfPipeVideoChannels.
•
Chapter 18, “Managing Nongraphic System Tasks,” describes clocks, memory
allocation, synchronous I/O, error handling and notification, and search paths.
•
Chapter 19, “Dynamic Data,” describes how to connect pfFlux, pfFCS, and
pfEngine nodes, which together can be used for animating geometries.
•
Chapter 20, “Active Surface Definition,” describes the Active Surface Definition
(ASD): a library that handles real-time surface meshing and morphing.
•
Chapter 21, “Light Points,” describes the calligraphic lights, which are intensely
bright lights.
•
Chapter 22, “Math Routines,” details the comprehensive math support provided as
part of OpenGL Performer.
•
Chapter 23, “Statistics,” discusses the various kinds of statistics you can collect and
display about the performance of your application.
•
Chapter 24, “Performance Tuning and Debugging,” explains how to use
performance measurement and debugging tools and provides hints for getting
maximum performance.
•
Chapter 25, “Building a Visual Simulation Application Using libpfv” describes a
modular approach to building an application using a graphical viewer.
•
Chapter 26, “Programming with C++,” discusses the differences between using the
C and C++ programming interfaces.
Sample Applications
You can find the sample code for all of the sample OpenGL Performer applications
installed under /usr/share/Performer/src/pguide on IRIX and Linux and under
%PFROOT%\Src\pguide on Microsoft Windows.
xlviii
007-1680-100
About This Guide
Conventions
This guide uses the following typographical conventions:
Bold
Used for function names, with parentheses appended to the name and
also for the names of window menus and buttons. Also, bold lowercase
letters represent vectors, and bold uppercase letters denote matrices.
Italics
Indicates variables, book titles, and glossary items.
Fixed-width
Used for filenames, operating system command names, command-line
option flags, code examples, and system output.
Bold Fixed-width
Indicates user input, such as items that you type in from the keyboard.
Note that in some cases it is convenient to refer to a group of similarly named
OpenGL Performer functions by a single name; in such cases an asterisk is used to
indicate all the functions whose names start the same way. For instance, pfNew*() refers
to all functions whose names begin with “pfNew”: pfNewChan(), pfNewDCS(),
pfNewESky(), pfNewGeode(), and so on.
Internet and Hardcopy Reading for the OpenGL Performer Series
The OpenGL Performer series include the followingmanuals in printed and online
formats:
•
OpenGL Performer Programmer’s Guide (this book)
•
OpenGL Performer Getting Started Guide
To read these online books, point your browser at the following:
•
http://docs.sgi.com
For general information about OpenGL Performer, use the following URL:
•
http://www.sgi.com/software/performer
The info-performer mailing list provides a forum for discussion of OpenGL
Performer including technical and nontechnical issues. Subscription requests should be
sent to [email protected] Much like the comp.sys.sgi.*
newsgroups on the Internet, it is not an official support channel but is monitored by
007-1680-100
xlix
About This Guide
several interested SGI employees familiar with the toolkit. The OpenGL Performer
mailing list archives are at the following URL:
•
http://oss.sgi.com/projects/performer/mail/info-performer/
Reader Comments
If you have comments about the technical accuracy, content, or organization of this
document, please tell us. Be sure to include the title and document number of the manual
with your comments. (Online, the document number is located in the front matter of the
manual. In printed manuals, the document number can be found on the back cover.)
You can contact us in any of the following ways:
•
Send e-mail to the following address:
[email protected]
•
Use the Feedback option on the Technical Publications Library World Wide Web
page:
http://docs.sgi.com
•
Contact your customer service representative and ask that an incident be filed in the
SGI incident tracking system.
•
Send mail to the following address:
Technical Publications
SGI
1600 Amphitheatre Pkwy., M/S 535
Mountain View, California 94043-1351
We value your comments and will respond to them promptly.
l
007-1680-100
Chapter 1
1. OpenGL Performer Programming Interface
This chapter describes the fundamental ideas behind the OpenGL Performer
programming interface in the following sections:
•
“General Naming Conventions” on page 1
•
“Class API” on page 3
•
“Base Classes” on page 6.
General Naming Conventions
The OpenGL Performer application programming interface (API) uses naming
conventions to help you understand what a given command will do and even predict the
appropriate names of routines for desired functionality. Following similar naming
practices in the software that you develop will make it easier for you and others on your
team to understand and debug your code.
The API is largely object-oriented; it contains classes of objects comprised of methods
that do the following:
•
Configure their parent objects.
•
Apply associated operations, based on the current configuration of the object.
Both C and C++ bindings are provided for OpenGL Performer. In addition, naming
conventions provide a consistent and predictable API and indicate the kind of operations
performed by a given command.
Prefixes
The prefix of the command tells you in which library a C command or C++ class is found.
All exposed OpenGL Performer base library C commands and C++ classes begin with
’pf’. The utility libraries use an additional prefix letter, such as ’pfu’ for the libpfutil
007-1680-100
1
1: OpenGL Performer Programming Interface
general utility library, ’pfi’ for the libpfui input handling library, and ’pfd’ for the
libpfdu database utility library. libpr-level commands still have the ’pf’ prefix as they
are still in the main libpf library
Header Files
Each OpenGL Performer library contains a main header file in
/usr/include/Performer on IRIX and Linux and in %PFROOT%\Include on
Microsoft Windows that contains type and class definitions, the C API for that library,
and global routines that are part of the C and C++ API. libpf is broken into two distinct
pieces: the low-level rendering layer, libpr, and the application layer, libpf, and each
has its own main header file: pr.h and pf.h. Since libpf is considered to include
libpr, pf.h includes pr.h. C++ class header files are found under the following
directories:
/usr/include/Performer/{pf, pr, ...} (IRIX and Linux)
%PFROOT%\Include\{pf, pr, ...} (Microsoft Windows)
Each class has its own C++ header file and that header must be included to use that class.
#include <Performer/pf.h>
#include <Performer/pf/pfGroup.h>
.....
pfGroup *group;
Naming in C and C++
All C++ class method names have an expanded C counterpart. Typically, the C routine
(function)will include the class name in the routine, whereas the C++ method will not.
C: pfGetPipeScreen();
C++: pipe->getScreen();
For some very general routines on the most abstract classes, the class name is omitted.
This is the case with the child API on pfNodes:
C: pfAddChild(node,child);
C++: node->addChild(child);
Command and type names are mixed case where the first letter of a new word in a name
is capitalized. C++ method names always start with a lower case letter.
2
007-1680-100
Class API
pfTexture *texture;
texture->loadFile();
Abbreviations
Type names do not use abbreviations. The C API acting on that type will often use
abbreviations for the type names, as will the associated tokens and enums.
In procedure names, a name will always be abbreviated or never, and the same
abbreviation will always be used and will be in the pfNew* C command. For example:
the pfTexture object uses ‘Tex’ in its API, such as pfNewTex(). If a type name has multiple
words, the abbreviation will use the first letter of the first words and then the first syllable
of the last word.
pfPipeWindow *pwin = pfNewPWin();
pfPipeVideoChannel *pvchan = pfNewPVChan();
pfTexLOD *tlod = pfNewTLOD();
Macros, Tokens, and Enums
Macros, tokens, and enums all use full upper-case. Token names associated with a class
and methods of a class start with the abbreviated name for that class, such as texture to
“tex” in PFTEX_SHARPEN.
Class API
The API of a given class, such as pfTexture, is comprised of the following:
007-1680-100
•
API to create an instance of the object
•
API to set parameters on the object
•
API to get those parameter settings
•
API to perform actions on the configured object
3
1: OpenGL Performer Programming Interface
Object Creation
Objects are always created with the following:
C: pfThing *thing = pfNewThing();
C++: pfThing *thing = new pfThing;
libpf objects are automatically created out of the shared memory arena. libpr objects
take as an argument an arena pointer which, if NULL, will cause allocation off the heap.
Set Routines
A set routine has the following form:
C: pfThingParam(thing, ... )
C++: thing->setParam()
Note that there is no ‘Set’ in the name in the C version.
Set routines are usually very fast and are not order dependent. Work required to process
the settings happens once when the object is first used after settings have changed. If
particularly expensive options must be done, there will be a pfConfigThing routine or
method to explicitly force this work that must be called before the object is to be used.
Get Routines
For every ‘set’ routine there is a matching ‘get’ routine to get back the value that was set.
C: pfGetThingParam(thing, ... )
C++: thing->getParam()
If the set/get is for a single value, that value is usually the return value of the routine. If
there are multiple values together, the ‘get’ routine will then take as arguments pointers
to result variables.
Getting Current In-Use Values
Get routines return values that have been previously set by the user, or default values if
no settings have been made. Sometimes a value other than the user-specified value is
currently in use and that is the value that you would like to get. For these cases, there is
a separate ‘GetCur’ routine to get the current in-use value.
4
007-1680-100
Class API
C:
pfGetCurThingParam()
C++: thing->getcurParam()
These ‘cur’ routines may only be able to give reasonable values in the process which
associated operations are happening. For example, to get the current texture
(pfGetCurTex()), you need to be in the draw process since that is the only process that
has a current texture.
Action Routines
An action routine has the following form:
C: pfVerbThing(), such as pfApplyTex()
C++: thing->verb(), such as tex->apply()
Action routines can have parameter scope and apply only to that parameter. These
routines have the following form
C: pfVerbThingParam(), such as pfApplyTexMinLOD()
C++: thing->verbParam(), such as tex->applyMinLOD()
Apply and Draw Routines
The Apply and Draw action routines do graphics operations and must happen either in
the draw process or in display list mode.
C: pfApplypfGState()
pfDrawGSet()
C++: gstate->apply()
gset->draw()
Enable and Disable of Modes
Features that can be enabled and disabled are done so with pfEnable() and pfDisable(),
respectively.
pfGetEnable() takes PFEN_* tokens naming the graphics state operation to enable or
disable. A GetEnable() is used to query enable status and will return 1 or 0 if the given
mode is enabled or disabled, respectively.
ex: pfEnable(PFEN_TEXTURE), pfDisable(PFEN_TEXTURE),
pfGetEnable(PFEN_TEXTURE);
007-1680-100
5
1: OpenGL Performer Programming Interface
Mode, Attribute, or Value
Classes instances are configured by having their internal fields set. These fields may be
simple modes or complex attribute structures. Mode values are ints or tokens, attributes
are typically pointers to objects, and values are floats.
pfGStateMode(gstate, PFSTATE_DECAL, PFDECAL_LAYER)
pfGStateAttr(gstate, PFSTATE_TEXTURE, texPtr)
pfGStateVal(gstate, PFSTATE_ALPHAREF, 0.5)
Base Classes
OpenGL Performer provides an object-oriented programming interface to most of its
data structures. Only OpenGL Performer functions can change the values of elements of
these data structures; for instance, you must call pfMtlColor() to set the color of a
pfMaterial structure rather than modifying the structure directly.
For a more transparent type of memory, OpenGL Performer provides pfMemory. All
object classes are derived from pfMemory. pfMemory instances must be explicitly
allocated with the new operator and cannot be allocated statically, on the stack, or
included directly in other object definitions. pfMemory is managed memory; it includes
special fields, such as size, arena, and ref count, that are initialized by the pfMemory
new() function.
Some very simple and unmanaged data types are not encapsulated for speed and easy
access. Examples include pfMatrix, pfSphere and pfVec3. These data types are referred
to as public structures and are inherited from pfStruct.
Unlike pfMemory, pfStructs can be handled as follows:
•
Allocated statically
•
Allocated on the stack
•
Included directly in other structure and object definitions
pfStructs allocated off the stack or allocated statically are not in the shared memory arena
and thus are not safe for multiprocessed use. Also, pfStructs allocated off the stack in a
procedure do not exist after the procedure exits so they should not be given to persistent
objects, such as a pfVec3 array of vertices for a pfGeoSet.
6
007-1680-100
Base Classes
In order to allow some functions to apply to multiple data types, OpenGL Performer uses
the concept of class inheritance. Class inheritance takes advantage of the fact that
different data types (classes) often share attributes. For example, a pfGroup is a node that
can have children. A pfDCS (Dynamic Coordinate System) has the same basic structure
as a pfGroup, but also defines a transformation to apply to its children—in other words,
the pfDCS data type inherits the attributes of the pfGroup and adds new attributes of its
own. This means that all functions that accept a pfGroup* argument will alternatively
accept a pfDCS* argument.
For example, pfAddChild() takes a pfGroup* argument, but appends child to the list of
children belonging to dcs:
pfDCS *dcs = pfNewDCS();
pfAddChild(dcs, child);
Because the C language does not directly express the notion of classes and inheritance,
arguments to functions must be cast before being passed, as shown in this example:
pfAddChild((pfGroup*)dcs, (pfNode*)child);
In the example above, no such casting is required because OpenGL Performer provides
macros that perform the casting when compiling with ANSI C, as shown in this example:
#define pfAddChild(g, c) pfAddChild((pfGroup*)g, (pfNode*)c)
Note: Using automatic casting eliminates type checking—the macros will cast anything
to the desired type. If you make a mistake and pass an unintended data type to a casting
macro, the results may be unexpected.
No such trickery is required when using the C++ API. Full type checking is always
available at compile time.
Inheritance Graph
The relations between classes can be arranged in a directed acyclic inheritance graph in
which each child inherits all of its parent’s attributes, as illustrated in Figure 1-1. OpenGL
Performer does not use multiple inheritance, so each class has only one parent in the
graph.
007-1680-100
7
1: OpenGL Performer Programming Interface
Note: It is important to remember that an inheritance graph is different from a scene
graph. The inheritance graph shows the inheritance of data elements and member
functions among user-defined data types; the scene graph shows the relationship among
instances of nodes in a hierarchical scene definition.
8
007-1680-100
Base Classes
pfObject
pfLight
pfPipe
pfMaterial
pfNode
pfGeoSet
pfChannel
pfFrustum
Some classes
found in libpf
Some classes
found in libpr
Figure 1-1
007-1680-100
Partial Inheritance Graph of OpenGL Performer Data Types
9
1: OpenGL Performer Programming Interface
OpenGL Performer objects are divided into two groups: those found in the libpf library
and those found in the libpr library. These two groups of objects have some common
attributes, but also differ in some respects.
While OpenGL Performer only uses single inheritance, some objects encapsulate others,
hiding the encapsulated object but also providing a functional interface that mimics its
original one. For example a pfChannel has a pfFrustum, a pfFrameStats has a pfStats, a
pfPipeWindow has a pfWindow, and a pfPipeVideoChannel has a pfVideoChannel. In
these cases, the first object in each pair provides functions corresponding to those of the
second. For example, pfFrustum has a routine:
pfMakeSimpleFrust(frust, 45.0f);
pfChannel has a corresponding routine:
pfMakeSimpleChan(channel, 45.0f);
libpr and libpf Objects
All of the major classes in OpenGL Performer are derived from the pfObject class. This
common, base class unifies the data types by providing common attributes and
functions. libpf objects are further derived from pfUpdatable. The pfUpdatable
abstract class provides support for automatic multibuffering for multiprocessing.
pfObjects have no special support for multiprocessing and so all processes share the
same copy of the pfObject in the shared arena. libpr objects allocated from the heap
are only visible in the process in which they are created or in child processes created after
the object. Changes made to such an object in one process are not visible in any other
process.
Explicit multibuffering of pfObjects is available through the pfFlux class. In general,
libpr provides lightweight and low-level modular pieces of functionality that are then
enhanced by more powerful libpf objects.
User Data
The primary attribute defined by the pfObject class is the custom data a user gets to
define on any pfObject called “user data.” pfUserDataSlot attaches the user-supplied
data pointer to user data. pfUserData attaches the user-supplied data pointer to user data
slot. Example 1-1 shows how to use user data.
10
007-1680-100
Base Classes
Example 1-1
How to Use User Data
typedef struct
{
float coeffFriction;
float density;
float *dataPoints;
}
myMaterial;
myMaterial
*granite;
granite = (myMaterial *)pfMalloc(sizeof(myMaterial), NULL);
granite->coeffFriction = 0.5f;
granite->density = 3.0f;
granite->dataPoints = (float *)pfMalloc(sizeof(float)*8, NULL);
graniteMtl = pfNewMtl(NULL);
pfUserData(graniteMtl, granite);
pfDelete() and Reference Counting
Most kinds of data objects in OpenGL Performer can be placed in a hierarchical scene
graph, using instancing when an object is referenced multiple times. Scene graphs can
become quite complex, which can cause problems if you are not careful. Deleting objects
can be a particularly dangerous operation, for example, if you delete an object that
another object still references.
Reference counting provides a bookkeeping mechanism that makes object deletion safe:
an object is never deleted if its reference count is greater than zero.
All libpr objects (such as pfGeoState and pfMaterial) have a reference count that
specifies how many other objects refer to it. A reference is made whenever an object is
attached to another using the OpenGL Performer routines shown in Table 1-1.
007-1680-100
11
1: OpenGL Performer Programming Interface
Table 1-1
Routines that Modify libpr Object Reference Counts
Routine
Action
pfGSetGState()
Attaches a pfGeoState to a pfGeoSet.
pfGStateAttr()
Attaches a state structure (such as a pfMaterial) to a pfGeoState.
pfGSetHlight()
Attaches a pfHighlight to a pfGeoSet.
pfTexDetail()
Attaches a detail pfTexture to a base pfTexture.
pfGSetAttr()
Attaches attribute and index arrays to a pfGeoSet.
pfTexImage()
Attaches an image array to a pfTexture.
pfAddGSet(),
Modify pfGeoSet/pfGeode association.
pfReplaceGSet(),
pfInsertGSet()
When object A is attached to object B, the reference count of A is incremented.
Additionally, if A replaces a previously referenced object C, then the reference count of
C is decremented. Example 1-2 demonstrates how reference counts are incremented and
decremented.
Example 1-2
Objects and Reference Counts
pfGeoState *gstateA, *gstateC;
pfGeoSet *gsetB;
/* Attach gstateC to gsetB. Reference count of gstateC
* is incremented. */
pfGSetGState(gsetB, gstateC);
/* Attach gstateA to gsetB, replacing gstateC. Reference
* count of gstateC is decremented and that of gstateA
* is incremented. */
pfGSetGState(gsetB, gstateA);
This automatic reference counting done by OpenGL Performer routines is usually all you
will ever need. However, the routines pfRef(), pfUnref(), and pfGetRef() allow you to
increment, decrement, and retrieve the reference count of a libpr object should you
wish to do so. These routines also work with objects allocated by pfMalloc().
12
007-1680-100
Base Classes
An object whose reference count is equal to 0 can be deleted with pfDelete(). pfDelete()
works for all libpr objects and all pfNodes but not for other libpf objects like pfPipe
and pfChannel. pfDelete() first checks the reference count of an object. If the reference
count is nonpositive, pfDelete() decrements the reference count of all objects that the
current object references, then it deletes the current object. pfDelete() does not stop here
but continues down all reference chains, deleting objects until it finds one whose count
is greater than zero. Once all reference chains have been explored, pfDelete returns a
boolean indicating whether it successfully deleted the first object or not. Example 1-3
illustrates the use of pfDelete() with libpr.
Example 1-3
Using pfDelete() with libpr Objects
pfGeoState *gstate0, *gstate1;
pfMaterial *mtl;
pfGeoSet *gset;
gstate0 = pfNewGState(arena); /* initial ref count is 0 */
gset = pfNewGSet(arena); /* initial ref count is 0 */
mtl = pfNewMtl(arena); /* initial ref count is 0 */
/* Attach mtl to gstate0. Reference count of mtl is
* incremented. */
pfGStateAttr(gstate0, PFSTATE_FRONTMTL, mtl);
/* Attach mtl to gstate1. Reference count of mtl is
* incremented. */
pfGStateAttr(gstate1, PFSTATE_FRONTMTL, mtl);
/* Attach gstate0 to gset. Reference count of gstate0 is
* incremented. */
pfGSetGState(gset, gstate0);
/* This deletes gset, gstate0, but not mtl since gstate1 is
* still referencing it. */
pfDelete(gset);
Example 1-4 illustrates the use of pfDelete() with libpf.
Example 1-4
Using pfDelete() with libpf Objects
pfGroup *group;
pfGeode *geode;
pfGeoSet *gset;
group = pfNewGroup(); /* initial parent count is 0 */
007-1680-100
13
1: OpenGL Performer Programming Interface
geode = pfNewGeode(); /* initial parent count is 0 */
gset = pfNewGSet(arena); /* initial ref count is 0 */
/* Attach geode to group. Parent count of geode is
* incremented. */
pfAddChild(group, geode);
/* Attach gset to geode. Reference count of gset is
* incremented. */
pfAddGSet(geode, gset);
/* This has no effect since the parent count of geode is 1.*/
pfDelete(geode);
/* This deletes group, geode, and gset */
pfDelete(group);
Some notes about reference counting and pfDelete():
•
All reference count modifications are locked so that they guarantee mutual
exclusion when multiprocessing.
•
Objects added to a pfDispList do not have their counts incremented due to
performance considerations.
•
In the multiprocessing environment of libpf, the successful deletion of a pfNode
does not have immediate effect but is delayed one or more frames until all processes
in all processing pipelines are through with the node. This accounts for the fact that
pfDispLists do not reference-count their objects.
•
pfUnrefDelete(obj) is shorthand for the following:
if(pfUnref(obj) ==0)
pfDelete(obj);
This is true when pfUnrefGetRef is atomic.
•
14
Objects whose count reaches zero are not automatically deleted by OpenGL
Performer. You must specifically request that an object be deleted with pfDelete()
or pfUnrefDelete().
007-1680-100
Base Classes
Copying Objects with pfCopy()
pfCopy() is currently implemented for libpr (and pfMalloc()) objects only. Object
references are copied and reference counts are modified appropriately, as illustrated in
Example 1-5.
Example 1-5
Using pfCopy()
pfGeoState *gstate0, *gstate1;
pfMaterial *mtlA, *mtlB;
gstate0 = pfNewGState(arena);
gstate1 = pfNewGState(arena);
mtlA = pfNewMtl(arena); /* initial ref count is 0 */
mtlB = pfNewMtl(arena); /* initial ref count is 0 */
/* Attach mtlA to gstate0. Reference count of mtlA is
* incremented. */
pfGStateAttr(gstate0, PFSTATE_FRONTMTL, mtlA);
/* Attach mtlB to gstate1. Reference count of mtlB is
* incremented. */
pfGStateAttr(gstate1, PFSTATE_FRONTMTL, mtlB);
/* gstate1 = gstate0. The reference counts of mtlA and mtlB
* are 2 and 0 respectively. Note that mtlB is NOT deleted
* even though its reference count is 0. */
pfCopy(gstate1, gstate0);
pfMalloc and the related routines provide a consistent method to allocate memory, either
from the user’s heap (using the C-library malloc() function) or from a shared memory
arena.
Printing Objects with pfPrint()
pfPrint() can print many different kinds of objects to a file; for example, you can print
nodes and geosets. To do so, you specify in the argument of the function the object to
print, the level of verbosity, and the destination file. An additional argument, which,
specifies different data according to the type of object being printed.
The different levels of verbosity include the following:
•
007-1680-100
PFPRINT_VB_OFF—no printing
15
1: OpenGL Performer Programming Interface
•
PFPRINT_VB_ON—minimal printing (default)
•
PFPRINT_VB_NOTICE—minimal printing (default)
•
PFPRINT_VB_INFO—considerable printing
•
PFPRINT_VB_DEBUG—exhaustive printing
If the object to print is a type of pfNode, which specifies whether the print traversal
should only traverse the current node (PFTRAV_SELF) or the entire scene graph where
the node specified in the argument is the root node (PFTRAV_SELF |
PFTRAV_DESCEND). For example, to print an entire scene graph, in which scene is the
root node, to the file, fp, with default verbosity, use the following line of code:
file = fopen (“scene.out”,”w”);
pfPrint(scene, PFTRAV_SELF | PFTRAV_DESCEND, PFPRINT_VB_ON, fp);
fclose(file);
If the object to print is a pfFrameStats, which should specify a bitmask of the frame
statistics classes that you want printed. The values for the bitmask include the following:
•
PFSTATS_ON enables the specified classes.
•
PFSTATS_OFF disables the specified classes.
•
PFSTATS_DEFAULT sets the specified classes to their default values.
•
PFSTATS_SET sets the class enable mask to enmask.
For example, to print select classes of a pfFrameStats structure, stats, to stderr, use the
following line of code:
pfPrint(stats, PFSTATS_ENGFX | PFFSTATS_ENDB |
PFFSTATS_ENCULL,PFSTATS_ON, NULL);
If the object to print is a pfGeoSet, which is ignored and information about that pfGeoSet
is printed according to the verbosity indicator. The output contains the types, names, and
bounding volumes of the nodes and pfGeoSets in the hierarchy. For example, to print the
contents of a pfGeoSet, gset, to stderr, use the following line of code:
pfPrint(gset, NULL, PFPRINT_VB_DEBUG, NULL);
Note: When the last argument, file, is set to NULL, the object is printed to stderr.
16
007-1680-100
Base Classes
Determining Object Type
Sometimes you have a pointer to a pfObject but you do not know what it really is—is it
a pfGeoSet, a pfChannel, or something else? pfGetType() returns a pfType which
specifies the type of a pfObject. This pfType can be used to determine the class ancestry
of the object. Another set of routines, one for each class, returns the pfType
corresponding to that class, for example, pfGetGroupClassType() returns the pfType
corresponding to pfGroup.
pfIsOfType() tells whether an object is derived from a specified type, as opposed to
being the exact type.
With these functions you can test for class type as shown in Example 1-6.
Example 1-6
General-Purpose Scene Graph Traverser
void
travGraph(pfNode *node)
{
if (pfIsOfType(node, pfGetDCSClassType()))
doSomethingTransforming(node);
/* If ’node’ is derived from pfGroup then recursively
* traverse its children */
if (pfIsOfType(node, pfGetGroupClassType()))
for (i = 0; i < pfGetNumChildren(node); i++)
travGraph(pfGetChild(node, i));
}
Because OpenGL Performer allows subclassing of built-in types, when decisions are
made based on the type of an object, it is usually better to use pfIsOfType() to test the
type of an object rather than to test for the strict equality of the pfTypes. Otherwise, the
code will not have reasonable default behavior with file loaders or applications that use
subclassing.
The pfType returned from pfGetType() is useful for programs but it is not in a readable
form for you. Calling pfGetTypeName() on a pfType returns a null-terminated ASCII
string that identifies an object’s type. For a pfDCS, for example, pfGetTypeName()
returns the string “pfDCS.” The type returned by pfGetType() can then be compared to
a class type using pfIsOfType(). Class types can be returned by methods such as
pfGetGroupClassType().
007-1680-100
17
Chapter 2
2. Setting Up the Display Environment
You can use the library libpf or libpfv as your base to build your application. For the
most part, this chapter (and guide) shows how to do so with libpf. For a more modular
approach using a graphical viewer, see Chapter 25, “Building a Visual Simulation
Application Using libpfv”.
The library libpf is a visual-database processing and rendering system. The visual
database has at its root a pfScene (as described in Chapter 3 and Chapter 4). The chain of
events necessary to proceed from the scene graph to the display includes the following:
1.
A pfScene is viewed by a pfChannel.
2. The pfChannel view of the pfScene is rendered by a pfPipe into a framebuffer.
3. A pfPipeWindow manages the framebuffer.
4. The images in the framebuffer are transmitted to a display system that is managed
by a pfPipeVideoChannel.
Figure 2-1 shows this chain of events.
007-1680-100
19
2: Setting Up the Display Environment
pfPipe
pfChannel 0
pfChannel 1
w
Windo
pfPipe
1
pfScene
w
Windo
pfPipe
0
Scene graph
Display system
pfChannel 0
Figure 2-1
pfChannel 1
From Scene Graph to Visual Display
The following sections describe how to implement this chain of events using pfPipes,
pfPipeWindows, and pfChannels directly or through the use of a configuration file:
20
•
“Using Pipes” on page 21
•
“Using Channels” on page 26
•
“Controlling the Video Output” on page 34
•
“Using Multiple Channels” on page 35
007-1680-100
Using Pipes
•
“Using Channel Groups” on page 40
•
“Importing OpenGL Multipipe SDK (MPK) Configuration Files” on page 44
Using Pipes
This section describes rendering pipelines (pfPipes) and their implementation in OpenGL
Performer. Each rendering pipeline draws into one or more windows (pfPipeWindows)
associated with a single geometry pipeline. A minimum of one rendering pipeline is
required, although it is possible to have more than one.
The Functional Stages of a Pipeline
This rendering pipeline comprises three primary functional stages:
APP
Simulation processing, which includes reading input from control
devices, simulating the vehicle dynamics of moving models, updating
the visual database, and interacting with other networked simulation
stations.
CULL
Traverses the visual database and determines which portions of it are
potentially visible (a procedure known as culling), selects a level of detail
(LOD) for each model, sorts objects and optimizes state management,
and generates a display list of objects to be rendered.
DRAW
Traverses the display list and issues graphics library commands to a
Geometry Pipeline in order to create an image for subsequent display.
Figure 2-2 shows the process flow for a single-pipe system. The application constructs
and modifies the scene definition (a pfScene) associated with a channel. The traversal
process associated with that channel’s pfPipe then traverses the scene graph, building an
OpenGL Performer libpr display list. As shown in the figure, this display list is used as
input to the draw process that performs the actual graphics library actions required to
draw the image.
007-1680-100
21
2: Setting Up the Display Environment
Application
Scene
Traversal/Cull
Draw
Frame Buffer
Pipeline 0
Figure 2-2
Single Graphics Pipeline
OpenGL Performer also provides additional processes for application processing tasks,
such as database loading and intersection traversals, but these processes are optinal and
are asynchronous to the software rendering pipeline(s).
An OpenGL Performer application renders images using one or more pfPipes. Each
pfPipe represents an independent software-rendering pipeline. Most IRIS systems
contain only one Geometry Pipeline; so, a single pfPipe is usually appropriate. This
single pipeline is often associated with a window that occupies the entire display surface.
Alternative configurations include Onyx3 systems with InfiniteReality3 graphics
(allowing up to 16 Geometry Pipelines). Applications can render into multiple windows,
each of which is connected to a single Geometry Pipeline through a pfPipe rendering
pipeline.
Figure 2-3 shows the process flow for a dual-pipe system. Notice both the differences and
similarities between these two figures. Each pipeline (pfPipe) is independent in
multiple-pipe configurations; the traversal and draw tasks are separate, as are the libpr
display lists that link them. In contrast, these pfPipes are controlled by the same
application process, and in many situations access the same shared scene definition.
22
007-1680-100
Using Pipes
Application
Pipeline 0
Figure 2-3
Scene
Pipeline 1
Traversal/Cull
Draw
Traversal/Cull
Draw
Frame Buffer
Frame Buffer
Dual Graphics Pipeline
Each of these stages can be combined into a single process or split into multiple processes
(pfMultiprocess) for enhanced performance on multiple CPU systems. Multiprocessing
and multiple pipes are advanced topics that are discussed in “Successful
Multiprocessing with OpenGL Performer” in Chapter 5.
Creating and Configuring a pfPipe
pfPipes and their associated processes are created when you call pfConfig(). They exist
for the duration of the application. After pfConfig(), the application can get handles to
the created pfPipes using pfGetPipe(). The argument to pfGetPipe() indicates which
pfPipe to return and is an integer between 0 and numPipes-1, inclusive. The pfPipe handle
is then used for further configuration of the pfPipe.
pfMultipipe() specifies the number of pfPipes desired; the default is one.
pfMultiprocess() specifies the multiprocessing mode used by all pfPipes. These two
routines are discussed further in “Successful Multiprocessing with OpenGL Performer”
in Chapter 5.
007-1680-100
23
2: Setting Up the Display Environment
A key part of pfPipe initialization is the determination of the graphics hardware pipeline
(or screen) and the creation of a window on that screen. The screen of a pfPipe can be set
explicitly using pfPipeScreen(). Under single pipe operation, pfPipes can also inherit the
screen of their first opened window. Under multipipe operation, the screen of all pfPipes
must be determined before the pipes are configured by pfConfigStage() or the first call
to pfFrame(). There may be other operations that require preset knowledge of the screen
even under single pipes, such as custom configuration of video channels, discussed in
“Creating and Configuring a pfChannel” on page 26.
Once the screen of a pfPipe has been set, it cannot be changed. All windows of a given
pfPipe must be opened on the same screen. A graphics window is associated with a
pfPipe through the pfPipeWindow mechanism. If you do not create a pfPipeWindow,
OpenGL Performer will automatically create and open a full screen window with a
default configuration for your pfPipe.
Once you create and initialize a pfPipe, you can query information about its
configuration parameters. pfGetPipeScreen() returns the index number of the hardware
pipeline for the pfPipe, starting from zero. On single-pipe systems the return value will
be zero. If no screen has been set, the return value will be (-1). pfGetPipeSize() returns
the full screen size, in pixels, of the rendering area associated with a pfPipe.
You may have application states associated with pfPipe stages and processes that need
special initialization. For this purpose, you may provide a stage configuration callback
for each pfPipe stage using pfStageConfigFunc(pipe, stageMask, configFunc) and
specify the pfPipe, the stage bitmask (including one or more of PFPROC_APP,
PFPROC_CULL, and PFPROC_DRAW), and your stage configuration callback routine.
At any time, you may call the function pfConfigStage() from the application process to
trigger the execution of your stage configuration callback in the process associated with
that pfPipe’s stage. The stage configuration callback will be invoked at the start of that
stage within the current frame (the current frame in the application process, and
subsequent frames through the cull and draw phases of the software rendering pipeline).
Use a pfStageConfigFunc() callback function to configure OpenGL Performer processes
not associated with pfPipes, such as the database process, PFPROC_DBASE, and the
intersection process, PFPROC_ISECT. A common process initialization task for real-time
applications is the selection and/or specification of a CPU on which to run.
24
007-1680-100
Using Pipes
Example of pfPipe Use
The sample source code shipped with OpenGL Performer includes several simple
examples of pfPipe use in both C and C++. Specifically, look at the following examples
under the C and C++ directories in /usr/share/Performer/src/pguide/libpf
for IRIX and Linux and in %PFROOT%\Src\pguide\libpfc for Microsoft Windows,
such as hello.c, simple.c, and multipipe.c.
Example 2-1 illustrates the basics of using pipes. The code in this example is adapted
from OpenGL Performer sample programs.
Example 2-1
pfPipes in Action
main()
{
int i;
/* Initialize OpenGL Performer */
pfInit();
/* Set number of pfPipes desired -- THIS MUST BE DONE
* BEFORE CALLING pfConfig().
*/
pfMultipipe(NumPipes);
/* set multiprocessing mode */
pfMultiprocess(PFMP_DEFAULT);
...
/* Configure OpenGL Performer and fork extra processes if
* configured for multiprocessing.
*/
pfConfig();
...
/* Optional custom mapping of pipes to screens.
* This is actually the reverse as the default.
*//
for (i=0; i < NumPipes; i++)
pfPipeScreen(pfGetPipe(i), NumPipes-(i+1));
{
/* set up optional DRAW pipe stage config callback */
pfStageConfigFunc(-1 /* selects all pipes */,
PFPROC_DRAW /* stage bitmask */,
ConfigPipeDraw /* config callback */);
/* Config func should be done next pfFrame */
pfConfigStage(i, PFPROC_DRAW);
007-1680-100
25
2: Setting Up the Display Environment
}
InitChannels();
...
/* trigger the configuration and opening of pfPipes
* and pfWindows
*/
pfFrame();
/* Application’s simulation loop */
while(!SimDone())
{
...
}
}
/* CALLBACK FUNCTIONS FOR PIPE STAGE INITIALIZATION */
void
ConfigPipeDraw(int pipe, uint stage)
{
/* Application state for the draw process can be initialized
* here. This is also a good place to do real-time
* configuration for the drawing process, if there is one.
* There is no graphics state or pfState at this point so no
* rendering calls or pfApply*() calls can be made.
*/
pfPipe *p = pfGetPipe(pipe);
pfNotify(PFNFY_INFO, PFNFY_PRINT,
“Initializing stage 0x%x of pipe %d”, stage, pipe);
}
Using Channels
This section describes how to use pfChannels. A pfChannel is a view of a scene. A
pfChannel is a required element for an OpenGL Performer application because it
establishes the visual frame of reference for what is rendered in the drawing process.
Creating and Configuring a pfChannel
When you create a new pfChannel, it is attached to a pfPipe for the duration of the
application. The pfPipe renders the pfScene viewed by the pfChannel into a
26
007-1680-100
Using Channels
pfPipeWindow that is managed by that pipe. Use pfNewChan() to create a new
pfChannel and assign it to a pfPipe. pfChannels are automatically assigned to the first
pfPipeWindow of the pfPipe. In the sample program, the following statement creates a
new channel and assigns it to pipe p.
chan = pfNewChan(p);
The pfChannel is automatically placed in the first pfPipeWindow of the pfPipe. A
pfPipeWindow is created automatically if one is not explicitly created with
pfNewPWin().
The simplest configuration uses one pipe, one channel, and one window. You can use
multiple channels in a single pfPipeWindow on a pfPipe, thereby allowing channels to
share hardware resources. Using multiple channels is an advanced topic that is discussed
in the section of this chapter on “Using Multiple Channels.” For now, focus your
attention on understanding the concepts of setting up and using a single channel.
The primary function of a pfChannel is to define the view of a scene. A view is fully
characterized by a viewport, a viewing frustum, and a viewpoint. The following sections
describe how to set up the scene and view for a pfChannel.
Setting Up a Scene
A pfChannel draws the pfScene set by pfChanScene(). A channel can draw only one
scene per frame but can change scenes from frame to frame. Other pfChannel attributes
such as LOD modifications, described in “pfLOD Nodes” in Chapter 3, affect the scene.
A pfChannel also renders an environmental model known as pfEarthSky. A pfEarthSky
defines the method for clearing the channel viewport before rendering the pfScene and
also provides environmental effects, including ground and sky geometry and fog and
haze. A pfEarthSky is attached to a pfChannel by pfChanESky().
Setting Up a Viewport
A pfChannel is rendered by a pfPipe into its pfPipeWindow. The screen area that displays
a pfChannel’s view is determined by the origin and size of the window and the channel
viewport specified by pfChanViewport. The channel viewport is relative to the lower left
corner of the window and ranges from 0 to 1. By default, a pfChannel viewport covers
the entire window.
007-1680-100
27
2: Setting Up the Display Environment
Suppose that you want to establish a viewport that is one-quarter of the size of the
window, located in the lower left corner of the window. Use pfChanViewport(chan, 0.0,
0.25, 0.0, 0.25) to set up the one-quarter window viewport for the channel chan.
You can then set up other channels to render to the other three-quarters of the window.
For example, you can use four channels to create a four-way view for an architectural or
CAD application. See “Using Multiple Channels” on page 35 to learn more about
multiple channels.
Setting Up a Viewing Frustum
A viewing frustum is a truncated pyramid that defines a viewing volume. Everything
outside this volume is clipped, while everything inside is projected onto the viewing
plane for display. A frustum is defined by the following:
•
field-of-view (FOV) in the horizontal and vertical dimensions
•
near and far clipping planes
A viewing frustum is created by the intersections of the near and far clipping planes with
the top, bottom, left, and right sides of the infinite viewing volume formed by the FOV
and aspect ratio settings. The aspect ratio is the ratio of the vertical and horizontal
dimensions of the FOV.
Figure 2-4 shows the parameters that define a symmetric viewing frustum. To establish
asymmetric frusta refer to the pfChannel(3pf) or pfFrustum(3pf) man pages for
further details.
28
007-1680-100
Using Channels
Horizontal FOV
x
Top
Far
Left
t
igh
Near
e
Lin
of s
y
Vertical FOV
Right
Bottom
Eyepoint
Aspect Ratio =
Figure 2-4
y
x
=
tan(vertical FOV/2)
tan(horizontal FOV/2)
Symmetric Viewing Frustum
The viewing frustum is called symmetric when the vertical half-angles are equal and the
horizontal half-angles are equal.
Field-of-View
The FOV is the angular width of view. Use pfChanFOV(chan, horiz, vert) to set up
viewing angles in OpenGL Performer. The quantities horiz and vert are the total
horizontal and vertical fields of view in degrees; usually you specify one and let OpenGL
Performer compute the other. If you are specifying one angle, pass any amount less than
or equal to zero, or greater than or equal to 180, as the other angle. OpenGL Performer
automatically computes the unspecified FOV angle to fit the pfChannel viewport using
the aspect-ratio preserving relationship
tan(vert/2) / tan(horiz/2) = aspect ratio
That is, the ratio of the tangents of the vertical and horizontal half-angles is equal to the
aspect ratio. For example, if horiz is 45 degrees and the channel viewport is twice as wide
as it is high (so the aspect ratio is 0.5), then the vertical field-of-view angle, vert, is
007-1680-100
29
2: Setting Up the Display Environment
computed to be 23.4018 degrees. If both angles are unspecified, pfChanFOV() assumes a
default value of 45 degrees for horiz and computes the value of vert as described.
Clipping Planes
Clipping planes define the near and far boundaries of the viewing volume. These
distances describe the extent of the visual range in the view, because geometry outside
these boundaries is clipped, meaning that it is not drawn.
Use pfChanNearFar(chan, near, far) to specify the distance along the line of sight from the
viewpoint to the near and far planes that bound the viewing volume. These clipping
planes are perpendicular to the line of sight. For the best visual acuity, choose these
distances so that near is as far away as possible from the viewpoint and far is as close as
possible to the viewpoint. Minimizing the range between near and far provides more
resolution for distance comparisons and fog computations.
Setting Up a Viewpoint
A viewpoint describes the position and orientation of the viewer. It is the origin of the
viewing location, the direction of the line of sight from the viewer to the scene being
viewed, and an up direction. The default viewpoint is at the origin (0, 0, 0) looking along
the +Y axis, with +Z up and +X to the right.
Use pfChanView(chan, point, dir) to define the viewpoint for the pfChannel identified by
chan. Specify the view origin for point in x, y, z world coordinates. Specify the view
direction for dir in degrees by giving the degree measures of the three Euler angles:
heading, pitch, and roll.
Heading is a rotation about the Z axis, pitch is a rotation about the X axis, and roll is a
rotation about the Y axis. The value of dir is the product of the rotations ROTy(roll) *
ROTx(pitch) * ROTz(heading), where ROTa(angle) is a rotation matrix about axis A of angle
degrees.
Angles have not only a degree value, but also a sense, + or –, indicating whether the
direction of rotation is clockwise or counterclockwise. Because different systems follow
different conventions, it is very important to understand the sense of the Euler angles as
they are defined by OpenGL Performer. OpenGL Performer follows the right-hand rule.
According to the right-hand rule, counterclockwise rotations are positive. This means
that a rotation about the X axis by +90 degrees shifts the +Y axis to the +Z axis, a rotation
30
007-1680-100
Using Channels
about the Y axis by +90 degrees shifts the +Z axis to the +X axis, and a rotation about the
Z axis by +90 degrees shifts the +X axis to the +Y axis.
Figure 2-5 shows a toy plane (somewhat reminiscent of the Ryan S-T) at the origin of a
coordinate system with the angles of rotation labeled for heading, pitch, and roll. The
arrows show the direction of positive rotation for each angle.
Z
+ Heading
Y
X
Figure 2-5
+ Roll
+ Pitch
Heading, Pitch, and Roll Angles
A roll motion tips the wings from side to side. A pitch motion tips the nose up or down.
Changing the heading, a yaw motion steers the plane. Accurate readings of these angles
are critical information for a pilot during a flight, and a thorough understanding of how
the angles function together is required for creation of an accurate flight simulation
visual with OpenGL Performer. The same is also true of marine and other vehicle
simulations.
Alternatively, you can use pfChanViewMat(chan, mat) to specify a 4x4 homogeneous
matrix mat that defines the view coordinate system for channel chan. The upper left 3x3
submatrix defines the coordinate system axes, and the bottom row vector defines the
origin of the coordinate system. The matrix must be orthonormal, or the results will be
undefined. You can construct matrices using tools in the libpr library.
007-1680-100
31
2: Setting Up the Display Environment
The origin and heading, pitch, and roll angles, or the view matrix, create a complete view
specification. The view specification can locate the eyepoint frame-of-reference origin at
any point in world coordinates. The gaze vector, the eye’s +Y axis, can point in any
direction. The up vector, the eye’s +Z axis, can point in any direction perpendicular to the
gaze vector.
You can query the system for the view and eyepoint-direction values with
pfGetChanView(), or obtain the view matrix directly with pfGetChanViewMat().
The view direction can be modified by one or more offsets, relative to the eyepoint
frame-of-reference. View offsets are useful in situations where several channels render
the same scene into adjacent displays for a wider field-of-view or higher resolution.
Offsets are also used for multiple viewer perspectives, such as pilot and copilot views.
Use pfChanViewOffsets(chan, xyz, hpr) to specify additional translation and rotation
offsets for the viewpoint and direction; xyz specifies a translation vector and hpr specifies
a heading/pitch/roll rotation vector. Viewing offsets are automatically added each
frame to the view direction specified by pfChanView() or pfChanViewMat().
For example, to create three different perspectives of the same scene as displayed by
three windows in an airplane cockpit, use azimuth offsets of 45, 0, and -45 for left,
middle, and right views. To create vertical view groups such as might be seen through
the windscreen of a helicopter, use both azimuth and elevation offsets. Once the view
offsets have been set up, you need only set the view once per frame. View offsets are
applied after the eyepoint position and gaze direction have been established. As with the
other angles, be aware that the conventions for measuring azimuth and elevation angles
vary between graphics systems; so, you should verify that the sense of the angles is
correct.
Example of Channel Use
Example 2-2 shows how to use various pfChannel-related functions. The code is derived
from OpenGL Performer sample programs.
Example 2-2
Using pfChannels
main()
{
pfInit();
...
pfConfig();
32
007-1680-100
Using Channels
...
InitScene();
InitPipe();
InitChannel();
/* Application main loop */
while(!SimDone())
{
...
}
}
void InitChannel(void)
{
pfChannel *chan;
chan = pfNewChan(pfGetPipe(0));
/* Set the callback routines for the pfChannel */
pfChanTravFunc(chan, PFTRAV_CULL, CullFunc);
pfChanTravFunc(chan, PFTRAV_DRAW, DrawFunc);
/* Attach the visual database to the channel */
pfChanScene(chan, ViewState->scene);
/* Attach the EarthSky model to the channel */
pfChanESky(chan, ViewState->eSky);
/* Initialize the near and far clipping planes */
pfChanNearFar(chan, ViewState->near, ViewState->far);
/* Vertical FOV is matched to window aspect ratio. */
pfChanFOV(chan, 45.0f/NumChans, -1.0f);
/* Initialize the viewing position and direction */
pfChanView(chan, ViewState->initView.xyz,
ViewState->initView.hpr);
}
/* CULL PROCESS CALLBACK FOR CHANNEL*/
/* The cull function callback. Any work that needs to be
* done in the cull process should happen in this function.
*/
void
CullFunc(pfChannel * chan, void *data)
{
007-1680-100
33
2: Setting Up the Display Environment
static long first = 1;
if (first)
{
if ((pfGetMultiprocess() & PFMP_FORK_CULL) &&
(ViewState->procLock & PFMP_FORK_CULL))
pfuLockDownCull(pfGetChanPipe(chan));
first = 0;
}
PreCull(chan, data);
pfCull();
/* Cull to the viewing frustum */
PostCull(chan, data);
}
/* DRAW PROCESS CALLBACK FOR CHANNEL*/
/* The draw function callback. Any graphics functionality
* outside OpenGL Performer must be done here.
*/
void
DrawFunc(pfChannel *chan, void *data)
{
PreDraw(chan, data);
/* Clear the viewport, etc. */
pfDraw();
/* Render the frame */
/* draw HUD, or whatever else needs
* to be done post-draw.
*/
PostDraw(chan, data);
}
Controlling the Video Output
Note: This is an advanced topic.
You use pfPipeVideoChannel to query and control the configuration of a hardware video
channel. The methods allow you to, for example, query or specify the origin and size of
the video output and scale the display.
34
007-1680-100
Using Multiple Channels
By default, all pfVideoChannels on a pfPipe use the first entire video channel on the
screen selected by the pfPipe. Each pfPipeWindow initially has a default
pfPipeVideoChannel already assigned to it. When pfChannels are added to
pfPipeWindows, they will be using, by default, this first pfPipeVideoChannel. You can
get a pfPipeVideoChannel of a pfPipeWindow with pfGetPWinPVChan() and
specifying the index of the pfPipeVideoChannel on the pfPipeWindow; the initial default
one will be at index 0. You can then reconfigure this pfPipeVideoChannel to select a
different video channel or change the attributes of the selected video channel. You can
create a pfPipeVideoChannel with pfNewPVChan(). To use this for a given pfChannel,
you must add it to a pfPipeWindow that will cover the screen area of the desired video
channel. When a pfPipeVideoChannel is added to a pfPipeWindow with
pfAddPWinPVChan(), the index into the pfPipeWindow list of video channels is
returned and by default the pfPipeVideoChannel will get the next active hardware video
channel after the previous pfPipeVideoChannel on that pfPipeWindow. You can
explicitly select the hardware video channel with pfPVChanId(). The pfChannel will
then reference this pfPipeVideoChannel through the index that you got back from
pfAddPWinPVChan() and assign to the pfChannel with pfChanPWinPVChanIndex().
pvc = pfNewPVChan(p);
pvcIndex = pfAddPWinPVChan(pw, pvc);
pfChanPWinPVChanIndex(chan, pvcIndex);
Note that the screen of the pfPipe must be known to fully specify the desired video
channel. Queries on the pfPipeVideoChannel will return values indicating unknown
configuration until the screen is known. The screen can be determined by OpenGL
Performer when the window is opened in the DRAW process but you can also explicitly
set the screen of the pfPipe with pfPipeScreen().
You can also get to the hardware video channel structure, pfVideoChannelInfo(), for
more configuration options, such as reading gamma data or even a specific video format.
For more information on pfPipeWindows and pfPipeVideoChannels, see Chapter 17,
“pfPipeWindows and pfPipeVideoChannels.”
Using Multiple Channels
Each rendering pipeline can render multiple channels with multiple
pfPipeVideoChannels to a single pfPipeWindows. Multiple pfPipeWindows can also be
used but at the cost of some additional processing overhead. The pfChannel is assigned
to the proper pfPipeWindow and selects its pfPipeVideoChannel from that
007-1680-100
35
2: Setting Up the Display Environment
pfPipeWindow. The pfChannel must also have a viewport, set with pfChanViewport(),
that covers the proper window area to match that of the desired pfPipeVideoChannel.
Each channel represents an independent viewpoint into either a shared or an
independent visual database. Different types of applications can have vastly different
pipeline-window-channel configurations. This section describes two extremes: visual
simulation applications, where there is typically one window per pipeline, and highly
interactive uses that require dynamic window and channel configuration.
One Window per Pipe, Multiple Channels per Window
Often there is a single channel associated with each pipeline, as shown in the top half of
Figure 2-6. This section describes two important uses for multiple-channel support—
multiple pipelines per system and multiple windows per pipeline—the second of which
is illustrated in the bottom half of Figure 2-6.
36
007-1680-100
Using Multiple Channels
Single Channel
Frame Buffer
Pipeline
Display
Device
Channel 0
Multiple Channel
Frame Buffer
Channel 0
Channel 1
Pipeline
Channel n-1
Display
Device
Figure 2-6
007-1680-100
Single-Channel and Multiple-Channel Display
37
2: Setting Up the Display Environment
One situation that requires multiple channels occurs when inset views must appear
within an image. A simple example of this application is a driving simulator in which the
screen image represents the view out the windshield. If a rear-view mirror is to be drawn,
it must overlay the main forward view to provide a separate view of the same database
within the borders of the simulated mirror’s frame.
Channels are rendered in the order that they are assigned to a pfPipeWindow on their
parent pfPipe. Channels, upon creation, are assigned to the end of the channel list of the
first window of their pfPipe. In the driving simulator example, creating pipes and
channels with the following structure creates two channels on a single shared pipeline:
pipeline = pfGetPipe(0);
frontView = pfNewChan(pipeline);
rearView = pfNewChan(pipeline);
In this case, OpenGL Performer’s actual drawing order becomes the following:
1.
Clear frontView.
2. Draw frontView.
3. Clear rearView.
4. Draw rearView.
This default ordering results in the rear-view mirror image always overlaying the
front-view image, as desired. You can control and reorder the drawing of channels within
a pfPipeWindow with the pfInsertChan(pwin, where, chan) and pfMoveChan(pwin,
where, chan) routines. More details about multiple channels and multiple window are
discussed in the next section.
When the host has multiple Geometry Pipelines, as supported on Onyx RealityEngine2
and InfiniteReality systems, you can create a pfPipe and pfChannel pair for each
hardware pipeline. The following code fragment illustrates a two-channel, two-pipeline
configuration:
leftPipe = pfGetPipe(0);
leftView = pfNewChan(leftPipe);
rightPipe = pfGetPipe(1);
rightView = pfNewChan(rightPipe);
This configuration forms the basis for a high-performance stereo display system, since
there is a hardware pipeline dedicated to each eye and rendering occurs in parallel.
38
007-1680-100
Using Multiple Channels
The two-channel stereo-view application described in this example and the inset-view
application described in the previous example can be combined to provide stereo views
for a driving simulator with an inset rear-view mirror. The correct management of each
eye’s viewpoint and the mirror reflection helps provide a convincing sense of physical
presence within the vehicle.
The third and most common multiple-channel situation involves support for multiple
video outputs per pipeline. To do this, first associate each pipeline with a set of
nonoverlapping channels, one for each desired view. Next, use one of the following
video-splitting methods:
•
Use the multi-channel hardware options, available from SGI, for systems such as the
8-channel Display Generator (DG) for InfiniteReality graphics, where you can create
up to eight independent video outputs from a single Graphics Pipeline, with each
video output corresponding to one of the tiled channels. The Octane video option
supports four video outputs and the RealityEngine2 MultiChannel Option supports
six video channels per Graphics Pipeline.
•
Connect multiple video monitors in series to a single pipeline’s video output.
Because each monitor receives the same display image, a masking bezel is used to
obscure all but the relevant portion of each display surface.
The three multiple-channel concepts described here can be used in combination. For
example, use of three InfiniteReality pipelines, each equipped with the 8-channel DG ,
allows creation of up to 24 independent video displays. The channel-tiling method can
also be used for some or all of these displays.
Example 2-3 shows how to use multiple channels on separate pipes.
Example 2-3
Multiple Channels, One Channel per Pipe
pfChannel *Chan[MAX_CHANS];
void InitChannel(int NumChans)
{
/* Initialize each channel on a separate pipe */
for (i=0; i< NumChans; i++)
Chan[i] = pfNewChan(pfGetPipe(i));
...
/* Make channel n/2 the master channel (can be any
* channel).
*/
007-1680-100
39
2: Setting Up the Display Environment
ViewState->masterChan = Chan[NumChans/2];
{
long share;
/* Get the default channel-sharing mask */
share = pfGetChanShare(ViewState->masterChan);
/* Add in the viewport share bit */
share |= PFCHAN_VIEWPORT;
if (GangDraw)
{
/* add GangDraw to channel share mask */
share |= PFCHAN_SWAPBUFFERS_HW;
}
pfChanShare(ViewState->masterChan, share);
}
/* Attach channels */
for (i=0; i< NumChans; i++)
if (Chan[i] != ViewState->masterChan)
pfAttachChan(ViewState->masterChan, Chan[i]);
...
/* Continue with channel initialization */
}
Using Channel Groups
In many multiple-channel situations, including the examples described in the previous
section, it is useful for channels to share certain attributes. For the three-channel cockpit
scenario, each pfChannel shares the same eyepoint while the left and right views are
offset using pfChanViewOffsets(). OpenGL Performer supports the notion of channel
groups, which facilitate attribute sharing between channels.
pfChannels can be gathered into channel groups that share like attributes. A channel
group is created by attaching one pfChannel to another, or to an existing channel group.
Use pfAttachChan() to create a channel group from two channels or to add a channel to
an existing channel group. Use pfDetachChan() to remove a pfChannel from a channel
group.
40
007-1680-100
Using Channel Groups
A channel share mask defines shared attributes for a channel group. The attribute tokens
listed in Table 2-1 are bitwise OR-ed to create the share mask.
Table 2-1
Attributes in the Share Mask of a Channel Group
Token
Shared Attributes
PFCHAN_FOV
Horizontal and vertical fields of view
PFCHAN_VIEW
View position and orientation
PFCHAN_VIEW_OFFSETS
(x, y, z) and (heading, pitch, roll) offsets of the view direction
PFCHAN_NEARFAR
Near and far clipping planes
PFCHAN_SCENE
All channels display the same scene.
PFCHAN_EARTHSKY
All channels display the same earth/sky model.
PFCHAN_STRESS
All channels use the same stress filter.
PFCHAN_LOD
All channels use the same LOD modifiers.
PFCHAN_SWAPBUFFERS
All channels swap buffers at the same time.
PFCHAN_SWAPBUFFERS_HW Synchronize swap buffers for channels on different graphics
pipelines.
Use pfChanShare() to set the share mask for a channel group. By default, channels share
all attributes except PFCHAN_VIEW_OFFSETS. When you add a pfChannel to a channel
group, it inherits the share mask of that group.
A change to any shared attribute is applied to all channels in a group. For example, if you
change the viewpoint of a pfChannel that shares PFCHAN_VIEW with its group, all
other pfChannels in the group will acquire the same viewpoint.
Two attributes are particularly important to share in adjacent-display multiple-channel
simulations: PFCHAN_SWAPBUFFERS and PFCHAN_LOD. PFCHAN_LOD ensures
that geometry that straddles displays is drawn the same way in each channel. In this case,
all channels will use the same LOD modifier when rendering their scenes so that LOD
behavior is consistent across channels. PFCHAN_SWAPBUFFERS ensures that channels
refresh the display with a new frame at the same time. pfChannels in different pfPipes
that share PFCHAN_SWAPBUFFERS_HW will frame-lock the graphics pipelines
together.
007-1680-100
41
2: Setting Up the Display Environment
Example 2-4 illustrates the use of multiple channels and channel sharing.
Example 2-4
Channel Sharing
pfChannel *Chan[MAX_CHANS];
main()
{
pfInit();
...
/* Set number of pfPipes desired.
* BEFORE CALLING pfConfig().
*/
pfMultipipe(NumPipes);
...
pfConfig();
...
InitScene();
THIS MUST BE DONE
InitChannels();
pfFrame();
/* Application main loop */
while(!SimDone())
{
...
}
}
void InitChannel(int NumChans)
{
/* Initialize all channels on pipe 0 */
for (i=0; i< NumChans; i++)
Chan[i] = pfNewChan(pfGetPipe(0));
...
/* Make channel n/2 the master channel (can be any
* channel).
*/
ViewState->masterChan = Chan[NumChans/2];
...
42
007-1680-100
Using Channel Groups
/* Attach all Channels as slaves to the master channel */
for (i=0; i< NumChans; i++)
if (Chan[i] != ViewState->masterChan)
pfAttachChan(ViewState->masterChan, Chan[i]);
pfSetVec3(xyz, 0.0f, 0.0f, 0.0f);
/* Set each channel’s viewing offset. In this case use
* many channels to create one multichannel contiguous
* frustum with a 45˚ field of view.
*/
for (i=0; i < NumChans; i++)
{
float fov = 45.0f/NumChans;
pfSetVec3(hpr, (((NumChans - 1) * 0.5f) - i) * fov,
0.0f, 0.0f);
pfChanViewOffsets(Chan[i], xyz, hpr);
}
...
/* Now, just configure the master channel and all of the
* other channels will share those attributes.
*/
chan = ViewState->masterChan;
pfChanTravFunc(chan, PFTRAV_CULL, CullFunc);
pfChanTravFunc(chan, PFTRAV_DRAW, DrawFunc);
pfChanScene(chan, ViewState->scene);
pfChanESky(chan, ViewState->eSky);
pfChanNearFar(chan, ViewState->near, ViewState->far);
pfChanFOV(chan, 45.0f/NumChans, -1.0f);
pfChanView(chan, ViewState->initView.xyz,
ViewState->initView.hpr);
...
}
Multiple Channels and Multiple Windows
For some interactive applications, you may want to be able to dynamically control the
configuration of channels and windows. OpenGL Performer allows you to dynamically
create, open, and close windows. You can also move channels among the windows of the
shared parent pfPipe, and reorder channels within a pfPipeWindow. Channels can be
007-1680-100
43
2: Setting Up the Display Environment
appended to the end of a pfPipeWindow channel list with pfAddChan() and removed
with pfRemoveChan(). A channel can only be attached to one pfPipeWindow — no
instancing of pfChannels is allowed. When a pfChannel is put on a pfPipeWindow, it is
automatically deleted from its previous pfPipeWindow. A channel that is not assigned to
a pfPipeWindow is not drawn (though it may still be culled).
You can control and reorder the drawing of channels within a pfPipeWindow with the
pfInsertChan(pwin, where, chan) and pfMoveChan(pwin, where, chan) routines. Both of
these routines do a type of insertion: pfInsertChan() will add chan to the pwin channel
list in front of the channel in the list at location where. pfMoveChan() will delete chan
from its old location and move it to where in the pwin channel list.
On IRIX systems, if you have pfChannels in different pfPipeWindows or pfPipes that are
supposed to combine to form a continuous scene, you will want to ensure that both the
vertical retrace and double buffering of these windows is synchronized. This is required
for both reasonable performance and visual quality. Use the genlock(7) system video
feature to ensure that the vertical retraces of different graphics pipelines are
synchronized. To synchronize double buffering, you want to either specify
PFCHAN_SWAPBUFFERS_HW in the share mask of the pfChannels and put the
pfChannels in a share group, or else create a pfPipeWindow swap group, discussed in
Chapter 17, “pfPipeWindows and pfPipeVideoChannels.”
Importing OpenGL Multipipe SDK (MPK) Configuration Files
OpenGL Multipipe SDK (MPK) is a software package for managing a multipipe
rendering environment. MPK uses a configuration file to describe the layout and
hierarchy of pipes, windows, and channels used by an application. The manual
SGI OpenGL Multipipe SDK User’s Guide describes the format of the configuration file.
An OpenGL Performer application can import MPK configuration files and skip the
explicit generation of pipes, windows, and channels. The library libpfmpk contains
functions for importing and configuring pipes, windows, and channels from an MPK
configuration file. The functions in libpfmpk store the display configuration
information in a pfvDisplayMngr class for easy access by the application. The
pfvDisplayMngr class is part of the pfvViewer implementation, which is described in
Chapter 25, “Building a Visual Simulation Application Using libpfv”.
The pfMPKImportFile() function takes an MPK configuration filename and generates
OpenGL Performer objects (pfPipes, pfPipeWindows, and pfChannels) accordingly. The
44
007-1680-100
Importing OpenGL Multipipe SDK (MPK) Configuration Files
function pfMPKImportConfig() is very similar. Instead of accepting a filename, it
accepts an MPK configuration class MPKConfig. The result of both these functions is
two-fold:
•
OpenGL Performer is configured with pipes, windows, and channels as specified in
the MPK configuration file.
•
The pfvDisplayMngr class contains a description of the configured display
topology (what pipe has what windows and what channels). It also contains
pointers to all the newly generated OpenGL Performer classes (pfPipe,
pfPipeWindow, and pfChannel).
The following is a code sample section for using the pfMPKImportFile() function:
// Initialize Performer
pfInit();
// Initialize the MultipipeSDK import library.
// No need to initialize MPK directly.
pfmpkInit();
// Import a MultipipeSDK file. This function calls pfConfig
// so we don’t have to.
pfMPKImportFile(config_filename);
// Load a model file for display.
pfNode *root = pfdLoadFile(model_filename);
// Attach loaded file to a new pfScene
pfScene *scene = new pfScene;
scene->addChild(root);
// Create a pfLightSource and attach it to scene
scene->addChild(new pfLightSource);
// Get access to the results of the MultipipeSDK import.
// pfvDisplayMngr contains pointers to all the
// pipes/windows/channels that the MultipipeSDK file specified.
pfvDisplayMngr *dm = pfvDisplayMngr::getMngr();
// All configured channels share the scene graph so we only
// have to assign one channel.
pfChannel *chan = dm -> getChan(0) -> getHandle();
chan->setScene(scene);
007-1680-100
45
2: Setting Up the Display Environment
Note: Since the pfvDisplayMngr class has no C API, you can only use libpfmpk from
C++ programs.
Figure 2-1 contains a diagram of the various objects participating in any libpfmpk
import operation.
pfPipe
pfChannel
pfPipeWindow
pfvDisplayMngr *dm
pfmpkImportFile()
MPK
config
file
Figure 2-7
dm
dm
dm
getPipe(i)
getPWin(i)
getChan(i)
getHandle()
getHandle()
getHandle()
Pointers to
pfPipe, pfPipeWindow,
pfChannel
The libpfmpk Import Operation
Both functions pfMPKImportFile() and pfMPKImportConfig() encapsulate the entire
OpenGL Performer configuration stage including the call to function pfConfig(). This
may be too inflexible for some applications. An additional set of functions in libpfmpk
provides lower-level access.
The following code sample shows the internal structure of function
pfMPKImportConfig(). All calls that pfMPKImportConfig() makes are publicly
accessible and an application can call them directly:
void pfMPKImportConfig(MPKConfig *cfg)
{
pfvDisplayMngr
*dm = pfvDisplayMngr::getMngr();
pfMPKImportInfo
info;
// Prepare temporary storage for pipe information.
info . numPipes = mpkConfigNPipes(cfg);
46
007-1680-100
Importing OpenGL Multipipe SDK (MPK) Configuration Files
info . pipeInfo = (pfMPKImportPipeInfo *)
malloc (info.numPipes * sizeof (pfMPKImportPipeInfo));
// Translate contents of MPKConfig into pfvDisplayMngr terms.
pfMPKPreConfig(cfg, &info);
// Let pfvDisplayMngr run all its pre-pfConfig processing.
dm -> preConfig();
// Performer configuration: After this point, we can start
// creating Performer windows and channels.
pfConfig();
// Inquire pipe sizes, and configure all pfvDisplayMngr
// objects that depend on them.
pfMPKPostConfig(cfg, &info);
// Ask pfDisplayMngr to create all the windows/channels.
dm -> postConfig();
// Invoke any pfPipe/pfPipeWindow/pfChannel calls that
// pfDisplayMngr doesn’t encapsulate.
pfMPKPostDMConfig(cfg, &info);
}
For completeness, the following is the source code for pfMPKImportFile():
void pfMPKImportFile(char *filename)
{
// Ask MPK to load the configuration file and pass to
// pfMPKImportConfig
pfMPKImportConfig(mpkConfigLoad(filename));
}
The function pfMPKPreConfig() traverses the MPKConfig class and creates its
pfvDisplayMngr equivalent. The function pfMPKPostConfig() patches the previous
pfvDisplayMngr configuration using pipe size information. This information becomes
available only after the call to pfConfig(); hence, patching cannot happen in
pfMPKPreConfig().
The function pfMPKPostDMConfig() traverses the pfvDisplayMngr hierarchy one last
time. This time pfvDisplayMngr already contains valid pointers to the
OpenGL Performer classes it creates. The function pfMPKPostDMConfig() makes
007-1680-100
47
2: Setting Up the Display Environment
OpenGL Performer calls on the pfPipe, pfPipeWindow, and pfChannel pointers. Since
pfvDisplayMngr does not encapsulate all configuration details, pfMPKPostDMConfig()
makes these configuration calls directly on the new OpenGL Performer classes.
48
007-1680-100
Chapter 3
3. Nodes and Node Types
A scene graph holds the data that defines a virtual world. The scene graph includes
low-level descriptions of object geometry and their appearance, as well as higher-level,
spatial information, such as specifying the positions, animations, and transformations of
objects, as well as additional application-specific data.
Scene graph data is encapsulated in many different types of nodes. One node might
contain the geometric data of an object; another node might contain the transformation
for that object to orient and position it in the virtual world. The nodes are associated in a
hierarchy that is an adirected, acyclic graph. OpenGL Performer and your application
can act on the scene graph to perform various complex operations efficiently, such as
database intersection and rendering scenes.
This chapter focuses on the data types themselves rather than instances of those types.
Chapter 4, “Database Traversal,” discusses traversing sample scene graphs in terms of
actual objects rather than abstract data types.
Nodes
A scene is represented by a graph of nodes. A node is a subclass of pfNode. Only nodes
can be in scene graphs and have child nodes. In general, nodes either contain descriptive
information about scene graph geometry, or they create groups and hierarchies of nodes.
Many classes, such as pfEngine and pfFlux, that are not nodes can interact with nodes.
Attribute Inheritance
The basic element of a scene hierarchy is the node. While OpenGL Performer supplies
many specific types of nodes, it also uses a concept called class inheritance, which allows
different node types to share attributes. An attribute is a descriptive element of geometry
or its appearance.
007-1680-100
49
3: Nodes and Node Types
pfNode
OpenGL Performer’s node hierarchy begins with the pfNode class, as shown in
Figure 3-1.
pfNode
pfGeode
pfText
pfGroup
pfASD
pfLightSource
pfBillboard
pfScene
pfPartition
pfLayer
pfLOD
pfFCS
Figure 3-1
pfSCS
pfSwitch
pfSequence
pfDCS
Nodes in the OpenGL Performer Hierarchy
All node types are derived from pfNode; they inherit pfNode’s attributes and the libpf
routines for setting and getting attributes. In general, a node type inherits the attributes
and routines of all its parent nodes in the type hierarchy.
50
007-1680-100
Nodes
Table 3-1 lists the basic node class and gives a simple description for each node type.
Table 3-1
OpenGL Performer Node Types
Node Type
Node Class
Description
pfNode
Abstract
Basic node type.
pfGroup
Branch
Groups zero or more children..
pfScene
Root
Parent of the visual database.
pfSCS
Branch
Static coordinate system.
pfDCS
Branch
Dynamic coordinate system.
pfFCS
Branch
Flux coordinate system.
pfDoubleSCS
Branch
Double-precision static coordinate system.
pfDoubleDCS
Branch
Double-precision dynamic coordinate system.
pfDoubleFCS
Branch
Double-precision flux coordinate system.
pfSwitch
Branch
Selects among multiple children.
pfSequence
Branch
Sequences through its children.
pfLOD
Branch
Level-of-detail node.
pfLayer
Branch
Renders coplanar geometry.
pfLightSource
Leaf
Contains specifications for a light source.
pfGeode
Leaf
Contains geometric specifications.
pfBillboard
Leaf
Rotates geometry to face the eyepoint.
pfPartition
Branch
Partitions geometry for efficient intersections.
pfText
Leaf
Renders 2D and 3D text.
pfASD
Leaf
Controls transition between LOD levels.
pfNode
As shown in Figure 3-1, all libpf nodes are arranged in a type hierarchy, which defines
the inheritance of functionality. A pfNode is an abstract class, meaning that a pfNode can
007-1680-100
51
3: Nodes and Node Types
never be explicitly created by an application, and all other nodes inherit the functionality
of pfNode. Its purpose is to provide a root to the type hierarchy and to define the
attributes that are common to all node types.
pfNode Attributes
The following pfNode attributes are inherited by all other libpf node types:
•
Node name
•
Parent list
•
Bounding geometry
•
Intersection and traversal masks
•
Callback functions and data
•
User data
Bounding geometry, intersection masks, user data, and callbacks are advanced topics
that are discussed in Chapter 4, “Database Traversal.”
The routines that set, get, and otherwise manipulate these attributes can be used by all
libpf node types, as indicated by the keyword ‘Node’ in the routine names. Nodes used
as arguments to pfNode routines must be cast to pfNode* to match parameter
prototypes, as shown in this example:
pfNodeName((pfNode*) dcs, "rotor_rotation");
However, you usually do not need to do this casting explicitly. When you use the C API
and compile with the –ansi flag (which is the usual way to compile OpenGL Performer
applications), libpf provides macro wrappers around pfNode routines that
automatically perform argument casting for you. When you use the C++ API, such type
casting is not necessary.
pfNode Operations
In addition to sharing attributes, certain basic operations are provided for all node types.
They include the following:
52
New
Create and return a handle to a new node.
Get
Get node attributes.
Set
Set node attributes.
007-1680-100
Nodes
Find
Find a node based on its name.
Print
Print node data.
Copy
Copy node data.
Delete
Delete a node.
The Set operation is implied in the node attribute name. The names of the
attribute-getting functions contain the string “Get.”
An Example of Scene Creation
Example 3-1 illustrates the creation of a scene that includes two different kinds of
pfNodes. (For information about pfScene nodes, see “pfScene Nodes” on page 61; for
information about pfDCS nodes, see “pfDCS Nodes” on page 62.)
Example 3-1
Making a Scene
pfScene *scene;
pfDCS *dcs1, *dcs2;
scene = pfNewScene();
dcs1 = pfNewDCS();
dcs2 = pfNewDCS();
pfCopy(dcs2, dcs1);
/* Create a new scene node */
/* Create a new DCS node */
/* Create a new DCS node */
/* Copy all node attributes */
/*
from dcs1 to dcs2 */
pfNodeName(scene, "Scene_Graph_Root"); /* Name scene node */
pfNodeName(dcs1,"DCS_1");
/* Name dcs1 */
pfNodeName(dcs2,"DCS_2");
/* Name dcs2 */
...
/* Use a pfGet*() routine to determine node name */
printf("Name of first DCS node is %s.", pfGetNodeName(dcs1));
...
/* Recursively free this node if it’s no longer referenced */
pfDelete(scene);
...
pfGroup
In addition to inheriting the pfNode attributes described in the “pfNode” section of this
chapter, a pfGroup also maintains a list of zero or more child nodes that are accessed and
007-1680-100
53
3: Nodes and Node Types
manipulated using group operators. Children of a pfGroup can be either branch or leaf
nodes. Traversals process the children of a pfGroup in left-to-right order.
Table 3-2 lists the pfGroup functions, with a description and a visual interpretation of
each.
Table 3-2
pfGroup Functions
Function Name
Description
pfAddChild(group, child)
Appends child to the list for group.
pfInsertChild(group, index, child)
Inserts child before the child whose
place in the list is index.
pfRemoveChild(group, child)
Detaches child from the list and
shifts the list to fill the vacant spot.
Returns 1 if child was removed.
Returns 0 if child was not found in
the list. Note that the “removed”
node is only detached, not deleted.
pfGetNumChildren(group)
Returns the number of children in
group.
Diagram
index = 2
4
The pfGroup nodes can organize a database hierarchy either logically or spatially. For
example, if your database contains a model of a town, a logical organization might be to
group all house models under a single pfGroup. However, this kind of organization is
less efficient than a spatial organization, which arranges geometry by location. A spatial
organization improves culling and intersection performance; in the example of the town,
spatial organization would consist of grouping houses with their local terrain geometry
instead of with each other. Chapter 4 describes how to spatially organize your database
for best performance.
54
007-1680-100
Nodes
The code fragment in Example 3-2 illustrates building a hierarchy using pfGroup nodes.
Example 3-2
Hierarchy Construction Using Group Nodes
scene = pfNewScene();
/* The following loop constructs a sample hierarchy by
* adding children to several different types of group
* nodes. Notice that in this case the terrain was broken
* up spatially into a 4x4 grid, and a switch node is used
* to cause only one vehicle per terrain node to be
* traversed.
*/
for(j = 0; j < 4; j++)
for(i = 0; i < 4; i++)
{
pfGroup *spatial_terrain_block = pfNewGroup();
pfSCS *house_offset = pfNewSCS();
pfSCS *terrain_block_offset = pfNewSCS();
pfDCS *car_position = pfNewDCS();
pfDCS *tank_position = pfNewDCS();
pfDCS *heli_position = pfNewDCS();
pfSwitch *current_vehicle_type;
pfGeode *heli, *car, *tank;
pfAddChild(scene, spatial_terrain_block);
pfAddChild(spatial_terrain_block,
terrain_block_offset);
pfAddChild(spatial_terrain_block, house_offset);
pfAddChild(spatial_terrain_block,
current_vehicle_type);
pfAddChild(current_vehicle_type, car_position);
pfAddChild(current_vehicle_type, tank_position);
pfAddChild(current_vehicle_type, heli_position);
pfAddChild(car_position, car);
pfAddChild(tank_position, tank);
pfAddChild(heli_position, heli);
}
...
/* The following shows how one might use OpenGL Performer to
* manipulate the scene graph at run time by adding and
* removing children from branch nodes in the scene graph.
007-1680-100
55
3: Nodes and Node Types
*/
for(j = 0; j < 4; j++)
for(i = 0; i < 4; i++)
{
pfGroup *this_terrain;
this_terrain = pfGetChild(scene, j*4 + i);
if (pfGetNumChildren(this_terrain) > 2)
this_tank = pfGetChild(this_terrain, 2);
if (is_tank_disable(this_tank))
{
pfRemoveChild(this_terrain, this_tank);
pfAddChild(disabled_tanks, this_tank);
}
}
...
Working with Nodes
This section describes the basic concepts involved in working with nodes. It explains
how shared instancing can be used to create multiple copies of an object, and how changes
made to a parent node propagate down to its children. A sample program that illustrates
these concepts is presented at the end of the chapter.
Instancing
A scene graph is typically constructed at application initialization time by creating and
adding new nodes to the graph. If a node is added to two or more parents it is termed
instanced and is shared by all its parents. Instancing is a powerful mechanism that saves
memory and makes modeling easier. libpf supports two kinds of instancing, shared
instancing and cloned instancing, which are described in the following sections.
Shared Instancing
Shared instancing is the result of simply adding a node to multiple parents. If an
instanced node has children, then the entire subgraph rooted by the node is considered
to be instanced. Each parent shares the node; thus, modifications to the instanced node
or its subgraph are experienced by all parents. Shared instances can be nested—that is,
an instance can itself instance other nodes.
56
007-1680-100
Working with Nodes
In the following sample code, group0 and group1 share a node:
pfAddChild(group0, node);
pfAddChild(group1, node);
Figure 3-2 shows the structure created by this code. Before the instancing operation, the
two groups and the node to be shared all exist independently, as shown in the left portion
of the figure. After the two function calls shown above, the two groups both reference the
same shared hierarchy. (If the original groups referenced other nodes, those nodes would
remain unchanged.) Note that each of the group nodes considers the shared hierarchy to
be its own child.
Group 1
Group 1
Group 0
Group 0
n
Figure 3-2
n
Shared Instances
Cloned Instancing
In many situations shared instancing is not desirable. Consider a subgraph that
represents a model of an airplane with articulations for ailerons, elevator, rudder, and
landing gear. Shared instances of the model result in multiple planes that share the same
articulations. Consequently, it is impossible for one plane to be flying with its landing
gear retracted while another is on a runway with its landing gear down.
Cloned instancing provides the solution to this problem by cloning—creating new copies
of variable nodes in the subgraph. Leaf nodes containing geometry are not cloned and
007-1680-100
57
3: Nodes and Node Types
are shared to save memory. Cloning the airplane model generates new articulation
nodes, which can be modified independently of any other cloned instance. The cloning
operation, pfClone(), is actually a traversal and is described in detail in Chapter 4.
Figure 3-3 shows the result of cloned instancing. As in the previous figure, the left half of
the drawing represents the situation before the operation, and the right half shows the
result of the operation.
G1
P1
G2
G1
G2
Root
P
D1
P2
Dynamic
coordinate
system
D
B
B
C
Figure 3-3
D2
A
A
Leaf
C
Cloned Instancing
The cloned instancing operation constructs new copies of each internal node of the
shared hierarchy, but uses the same shared instance of all the leaf nodes. In use, this is an
important distinction, because the number of internal nodes may be relatively few, while
the number and content of geometry-containing leaf nodes is often quite extensive.
Nodes G1 and G2 in Figure 3-3 are the groups that form the root nodes after the cloned
instancing operation is complete. Node P is the parent or root node of the instanced
object, and D is a dynamic coordinate system contained within it. Nodes A, B, and C are
the leaf geometry nodes; they are shared rather than copied.
The code in Example 3-3 shows how to create cloned instances.
58
007-1680-100
Working with Nodes
Example 3-3
Creating Cloned Instances
pfGroup *g1, *g2, *p;
pfDCS *d;
pfGeode *a, *b, *c;
...
/* Create initial instance of scene hierarchy of p under
* group g1: add a DCS to p, then add three pfGeode nodes
* under the DCS.
*/
pfAddChild(g1,p);
pfAddChild(p,d);
pfAddChild(d,a);
pfAddChild(d,b);
pfAddChild(d,c);
...
/* Create cloned instance version of p under g2 */
pfAddChild(g2, pfClone(p,0));
/* Notice that pfGeodes are cloned by instancing rather than
* copying. Also notice that the second argument to
* pfClone() is 0; that argument is currently required by
* OpenGL Performer to be zero.
*/
...
Bounding Volumes
The libpf library uses bounding volumes for culling and to improve intersection
performance. libpf computes bounding volumes for all nodes in a database hierarchy
unless the bound is explicitly set by the application. The bounding volume of a branch
node encompasses the spatial extent of all its children. libpf automatically recomputes
bounds when children are modified.
By default, bounding volumes are dynamic; that is, libpf automatically recomputes
them when children are modified. For instance, in Example 3-4 when the DCS is rotated,
nothing more needs to be done to update the bounding volume for g1.
Example 3-4
Automatically Updating a Bounding Volume
pfAddChild(g1,dcs);
pfAddChild(dcs, helicopter);
007-1680-100
59
3: Nodes and Node Types
...
pfDCSRot(dcs, heading+10.0f, pitch,roll);
...
pfDCSRot(dcs, heading, pitch - 5.0f, roll + 2.0f);
In some cases, you may not want bounding volumes to be recomputed automatically. For
example, in a merry-go-round with horses moving up and down, you know that the
horses stay within a certain volume. Using pfNodeBSphere(), you can specify a
bounding sphere within which the horse always remains and tell OpenGL Performer
that the bounding volume is “static”—not to be updated no matter what happens to the
node’s children. You can always force an update by setting the bounding volume to
NULL with pfNodeBSphere(), as follows:
pfNodeBSphere(node, NULL, NULL, PFBOUND_STATIC);
At the lowest level, within pfGeoSets, bounding volumes are maintained as
axially-aligned boxes. When you add a pfGeoSet to a pfGeode or directly invoke
pfGetGSetBBox() on the pfGeoSet, a bounding box is created for the pfGeoSet. Neither
the bounding box of the pfGeoSet nor the bounding volume of the pfGeode is updated
if the geometry changes inside the pfGeoSet. You can force an update by setting the
pfGeoSet bounding box and then the pfGeode bounding volume to a NULL bounding
box, as follows:
•
Recompute the pfGeoSet bounding box from the internal geometry:
pfGSetBBox(gset, NULL);
•
Recompute the pfGeode bounding volume from the bounding boxes of its
pfGeoSets:
pfNodeBSphere(geode, NULL, PFBOUND_DYNAMIC);
60
007-1680-100
Node Types
Node Types
This section describes the node types and the functions for working with each node type.
pfScene Nodes
A pfScene is a root node that is the parent of a visual database. Use pfNewScene() to
create a new scene node. Before the scene can be drawn, you must call
pfChanScene(channel, scene) to attach it to a pfChannel.
Any nodes that are within the graph that is parented by a pfScene are culled and drawn
once the pfScene is attached to a pfChannel. Because pfScene is a group, it uses pfGroup
routines; however, a pfScene cannot be the child of any other node. The following
statement adds a pfGroup to a scene:
pfAddChild(scene,root);
In the simplest case, the pfScene is the only node you need to add. Once you have a
pfPipe, pfChannel, and pfScene, you have all the necessary elements for generating
graphics using OpenGL Performer.
pfScene Default Rendering State
The pfScene nodes may specify a global pfGeoState that all other pfGeoStates in nodes
below the pfScene will inherit from. Specification of this scene pfGeoState is done via the
function pfSceneGState(). This functionality allows for the subtle optimization of
pushing the most frequently used pfGeoState attributes for a particular scene graph into
a global state and having the individual states inherit these attributes rather than specify
them. This can save OpenGL Performer work during culling (by having to ‘unwrap’
fewer pfGeoStates) and thus possibly increase frame rate.
There are several database utility functions in libpfdu designed to help with this
optimization. pfdMakeSceneGState() returns an ‘optimal’ pfGeoState based on a list of
pfGeoStates. pfdOptimizeGStateList() takes an existing global pfGeoState, a new global
pfGeoState, and a list of pfGeoStates that should be optimized and cause all attributes of
pfGeoStates in the list of pfGeoStates to be inherited if they are the same as the attribute
in the new global pfGeoState. Lastly, pfdMakeSharedScene() causes this optimization to
happen for all of the pfGeoStates under the pfScene that was passed into the function.
For more information on pfGeoStates see Chapter 8, “Geometry,” which discusses
libpr in more detail. For more information on the creation and optimization of
007-1680-100
61
3: Nodes and Node Types
databases, see Chapter 7, “Importing Databases,” which discusses building database
converters and libpfdu.
pfSCS Nodes
A pfSCS is a branch node that represents a static coordinate system. A pfSCS node
contains a fixed modeling transformation matrix that cannot be changed once it is
created. pfSCS nodes are useful for positioning models within a database. For example,
a house that is modeled at the origin should be placed in the world with a pfSCS because
houses rarely move during program execution.
Use pfNewSCS(matrix) to create a new pfSCS using the transformation defined by matrix.
To find out what matrix was used to create a given pfSCS, call pfGetSCSMat().
For best graphics performance, matrices passed to pfSCS nodes (and the pfDCS node
type described in the next section) should be orthonormal (translations, rotations, and
uniform scales). Nonuniform scaling requires renormalization of normals in the graphics
pipe. Projections and other non-affine transformations are not supported.
While pfSCS nodes are useful in modeling, using too many of them can reduce culling,
rendering, and intersection performance. For this reason, libpf provides the pfFlatten()
traversal. pfFlatten() will traverse a scene graph and apply static transformations
directly to geometry to eliminate the overhead associated with managing the
transformations. pfFlatten() is described in detail in Chapter 4, “Database Traversal.”
pfDCS Nodes
A pfDCS is a branch node that represents a dynamic coordinate system. Use a pfDCS
when you want to apply an initial transformation to a node and also change the
transformation during the application. Use a pfDCS to articulate moving parts and to
show object motion.
Use pfNewDCS() to create a new pfDCS. The initial transformation of a pfDCS is the
identity matrix. Subsequent transformations are set by specifying a new transformation
matrix, or by replacing the rotation, scale, or translation in the current transformation
matrix. The pfDCS transforms each child C(i) to C(i)∗Scale∗Rotation∗Translation.
62
007-1680-100
Node Types
Table 3-3 lists functions for manipulating a pfDCS, including rotating, scaling, and
translating the children of the pfDCS.
Table 3-3
pfDCS Transformations
Function Name
Description
pfNewDCS()
Create a new pfDCS node.
pfDCSTrans()
Set the translation coordinates to x, y, z.
pfDCSRot()
Set the rotation transformation to h, p, r.
pfDCSCoord()
Rotate and translate by coord.
pfDCSScale()
Scale by a uniform scale factor.
pfDCSMat()
Use a matrix for transformations.
pfGetDCSMat()
Retrieve the current matrix for a given pfDCS.
pfFCS Nodes
A pfFCS is a branch node that represents a flux coordinate system. The transformation
matrix of a pfFCS is contained in the pfFlux which is linked to it. This linkage allows a
pfEngine to animate the matrix of a pfFCS. The linkage also allows multiple pfFCSs to
share the same transformation.
Use pfNewFCS(flux) to create a new pfFCS linked to flux.
Table 3-4 lists functions for manipulating a pfFCS. pfFCS, pfFlux, and pfEngine are fully
described in Chapter 19, “Dynamic Data.”
Table 3-4
007-1680-100
pfFCS Functions
Function
Description
pfNewFCS()
Create a new pfFCS node.
pfFCSFlux()
Link a flux to a given pfFCS.
pfGetFCSFlux()
Get a pointer to the flux linked to a given pfFCS.
63
3: Nodes and Node Types
Table 3-4
pfFCS Functions (continued)
Function
Description
pfGetFCSMat()
Retrieve the current matrix for a given pfFCS.
pfGetFCSMatPtr()
Get a pointer to the current matrix for a given pfFCS.
pfDoubleSCS Nodes
The pfDoubleSCS nodes are double-precision versions of pfSCS nodes. Instead of storing
a pfMatrix, they store a pfMatrix4d, a 4x4 matrix of double-precision numbers.
See the section “pfDoubleDCS Nodes” for a discussion on using double-precision matrix
nodes.
pfDoubleDCS Nodes
pfDoubleDCS nodes are double-precision versions of pfDCS nodes. Instead of a
pfMatrix, they maintain a pfMatrix4d, a 4x4 matrix of double-precision numbers.
Double-precision nodes are useful for modeling and rendering objects very far from the
origin of the database. The following example demonstrates how double-precision nodes
help. Consider a model of the entire Earth and visualize a model of a car moving on the
surface of the Earth. Placing the origin of the Earth model in the center of the Earth makes
the car object on the surface of the Earth very far from the origin. In Figure 3-4, the
distance from the center of the Earth to the car or to the camera is larger than D, and the
distance from the viewer to the car is d. D is very large; therefore, single-precision floating
point numbers cannot express small changes in the car position. The motion of the car
will be shaky and unsmooth.
One potential solution for the shaky car motion is to use double-precision matrices in
OpenGL. Unfortunately, the underlying hardware implementation does not support
double-precision values. All values are converted to single-precision floating point
numbers and OpenGL Performer cannot eliminate the shaky motion.
64
007-1680-100
Node Types
Eye
d
Car
D
Origin (0,0,0)
Earth
Figure 3-4
A Scenario for Using Double-Precision Nodes
In order to solve the shaky motion problem, we observe the following: we usually want
to see small translations of an object when the camera is fairly close to that object. If we
look at the car from 200 miles away, we do not care to see a 10-inch translation in its
position. Therefore, if we could dynamically drag the origin with the camera, then any
object will be close enough to the origin when the camera is near it, which is exactly when
we want to see its motion smoothly.
Double-precision matrix nodes (pfDoubleSCS, pfDoubleDCS, and pfDoubleFCS) allow
modeling with a dynamic origin. We start by setting the pfChannel viewing matrix to the
identity matrix. This puts the channel eyepoint in the origin. We create a scene graph as
in Figure 3-5. Each pfGeode represents a tile of the Earth surface. We model each tile with
a local origin somewhere within the tile.
Each of the pfDoubleDCS nodes above the pfGeode nodes contains a transformation that
sends the node under it to its correct position around the globe. We set the transformation
in the pfDoubleDCS node marked EYE to the inverse of the matrix taking an object to the
true camera position. This transforms all nodes under EYE to a coordinate system with
the eyepoint in the origin.
007-1680-100
65
3: Nodes and Node Types
pfScene
EYE pfDoubleDCS
pfDoubleDCS # 0
pfGeode # 0
Figure 3-5
pfDoubleDCS # 1
pfGeode # 1
....
....
pfDoubleDCS # N
pfGeode # N
pfDoubleDCS Nodes in a Scene Graph
In more practical terms, we set the channel camera position to the origin with the
following call:
pfChanViewMat(chan, pfIdentMat);
The following code fragment loads the EYE pfDoubleDCS node with the correct matrix.
We call the function with EYE as the first parameter and the camera position in the
second parameter:
void
loadViewingMatrixOnDoubleDCS (pfDoubleDCS *ddcs, pfCoordd *coord)
{
pfMatrix4d
mat, invMat;
pfMakeCoorddMat4d (mat, coord);
pfInvertOrthoNMat4d (invMat, mat);
pfDoubleDCSMat (ddcs, invMat);
}
66
007-1680-100
Node Types
pfDoubleFCS Nodes
A pfDoubleFCS node is similar to a pfFCS node. Instead of a single-precision matrix, it
maintains a pfFlux with a double-precision matrix. See “pfDoubleDCS Nodes” for
information on using pfDoubleFCS nodes.
pfSwitch Nodes
A pfSwitch is a branch node that selects one, all, or none of its children. Use
pfNewSwitch() to return a handle to a new pfSwitch. To select all the children, use the
PFSWITCH_ON argument to pfSwitchVal(). Deselect all the children (turning the
switch off) using PFSWITCH_OFF. To select a single child, give the index of the child
from the child list. To find out the current value of a given switch, call pfGetSwitchVal().
Example 3-5 (in the “pfSequence Nodes” section) illustrates a use of pfSwitch nodes to
control pfSequence nodes.
pfSequence Nodes
A pfSequence is a pfGroup that sequences through a range of its children, drawing each
child for a specified duration. Each child in a sequence can be thought of as a frame in an
animation. A sequence can consist of any number of children, and each child has its own
duration. You can control whether an entire sequence repeats from start to end, repeats
from end to start, or terminates.
Use pfNewSeq() to create and return a handle to a new pfSequence. Once the
pfSequence has been created, use the group function pfAddChild() to add the children
that you want to animate.
Table 3-5 describes the functions for working with pfSequences.
Table 3-5
007-1680-100
pfSequence Functions
Function
Description
pfNewSeq()
Create a new pfSequence node.
pfSeqTime()
Set the length of time to display a frame.
pfGetSeqTime()
Find out the time allotted for a given frame.
67
3: Nodes and Node Types
Table 3-5
pfSequence Functions (continued)
Function
Description
pfSeqInterval()
Set the range of frames and sequence type.
pfGetSeqInterval()
Find out interval parameters.
pfSeqDuration()
Control the speed and number of repetitions of the entire sequence.
pfGetSeqDuration() Retrieve speed and repetition information for the sequence.
pfSeqMode()
Start, stop, pause, and resume the sequence.
pfGetSeqMode()
Find out the sequence’s current mode.
pfGetSeqFrame()
Get the current frame.
Example 3-5 demonstrates a possible use of both switches and sequences. First,
sequences are set up to contain animation sequences for explosions, fire, and smoke; then
a switch is used to control which sequences are currently active.
Example 3-5
Using pfSwitch and pfSequence Nodes
pfSwitch *s;
pfSequence *explosion1_seq, *explosion2_seq, *fire_seq,
*smoke_seq;
...
s = pfNewSwitch();
explosion1_seq = pfNewSeq();
explosion2_seq = pfNewSeq();
fire_seq = pfNewSeq();
smoke_seq = pfNewSeq();
pfAddChild(s, explosion1_seq);
pfAddChild(s, explosion2_seq);
pfAddChild(s, fire_seq);
pfAddChild(s, smoke_seq);
pfSwitchVal(s, PFSWITCH_OFF);
...
if (direct_hit)
{
pfSwitchVal(s, PFSWITCH_ON); /* Select all sequences */
/* Set first explosion sequence to go double normal
* speed and repeat 3 times. */
68
007-1680-100
Node Types
pfSeqMode(explosion1_seq, PFSEQ_START);
pfSeqDuration(explosion1_seq, 2.0f, 3);
/* Set second explosion sequence to display first child
* of sequence for 2 seconds before continuing. */
pfSeqMode(explosion2_seq, PFSEQ_START);
pfSeqTime(explosion2, 0.0f, 2.0f);
/* Set fire to wait on first frame of sequence until .3
* seconds after second explosion. */
pfSeqMode(fire_seq, PFSEQ_START);
pfSeqTime(fire_seq, 0.0f, 2.3f);
/* Set smoke to wait until .1 seconds after fire. */
pfSeqMode(smoke_seq, PFSEQ_START);
pfSeqTime(smoke_seq, 0.0f, 2.4f);
}
else if (explosion && (expl_type == 0))
{
pfSeqMode(explosion1_seq, PFSEQ_START);
pfSwitchVal(s, 0);
}
else if (explosion && (expl_type == 1))
{
pfSeqMode(explosion2_seq, PFSEQ_START);
pfSwitchVal(s, 1);
}
else if (fire_is_burning)
{
pfSeqMode(fire_seq, PFSEQ_START);
pfSwitchVal(s, 2);
}
else if (smoking)
{
pfSeqMode(smoke_seq, PFSEQ_START);
pfSwitchVal(s, 3);
}
else
pfSwitchVal(s, PFSWITCH_OFF);
...
007-1680-100
69
3: Nodes and Node Types
pfLOD Nodes
A pfLOD is a level-of-detail node. Level-of-detail switching is an advanced concept that
is discussed in Chapter 5, “Frame and Load Control.” A level-of-detail node specifies
how its children are to be displayed, based on the visual range from the channel’s
viewpoint. Each child has a defined range, and the entire pfLOD has a defined center.
Table 3-6 describes the functions for working with pfLODs.
Table 3-6
pfLOD Functions
Function
Description
pfNewLOD()
Create a level of detail node.
pfLODRange()
Set a range at which to use a specified child node.
pfGetLODRange()
Find out the range for a given node.
pfLODCenter()
Set the pfLOD center.
pfGetLODCenter()
Retrieve the pfLOD center.
pfLODTransition()
Set the width of a specified transition.
pfGetLODTransition() Get the width of a specified transition.
pfASD Nodes
The pfASD nodes handle dynamic generation and morphing of the visible part of a
surface based on multiple LODs. pfASD nodes allow for the smooth LOD transition of
large and complex surfaces, such as large area terrain. For information on pfASD nodes,
see Chapter 20, “Active Surface Definition.”
pfLayer Nodes
A pfLayer is a leaf node that resolves the visual priority of coplanar geometry. A pfLayer
allows the application to define a set of base geometry and a set of layer geometry
(sometimes called decal geometry). The base geometry and the decal geometry should be
coplanar, and the decal geometry must lie within the extent of the base polygons.
70
007-1680-100
Node Types
Table 3-7 describes the functions for working with pfLayers.
Table 3-7
pfLayer Functions
Function
Description
pfNewLayer()
Create a pfLayer node.
pfLayerMode()
Specify a hardware mode to use in drawing decals.
pfGetLayerMode()
Get the current mode.
pfLayerBase()
Specify the child containing base geometry.
pfGetLayerBase()
Find out which child contains base geometry.
pfLayerDecal()
Specify the child containing decal geometry.
pfGetLayerDecal()
Find out which child contains decal geometry.
The pfLayer nodes can be used to overlay any sort of markings on a given polygon and
are important to avoid flimmering. Example 3-6 demonstrates how to display runway
markings as a decal above a coplanar runway. This example uses the performance mode
PFDECAL_BASE_FAST for layering; as described in the pfLayer and pfDecal man
pages, other available modes are PFDECAL_BASE_HIGH_QUALITY,
PFDECAL_BASE_DISPLACE, and PFDECAL_BASE_STENCIL.
Example 3-6
Marking a Runway with a pfLayer Node
pfLayer *layer;
pfGeode *runway, *runway_markings;
...
/* avoid flimmering of runway and runway_markings */
layer = pfNewLayer();
pfLayerBase(layer, runway);
pfLayerDecal(layer, runway_markings);
pfLayerMode(layer, PFDECAL_BASE_FAST);
pfGeode Nodes
The pfGeode node is short for geometry node and is the primary node for defining
geometry in libpf. A pfGeode contains a list of geometry structures called pfGeoSets,
which are part of the OpenGL Performer libpr library. pfGeoSets encapsulate graphics
007-1680-100
71
3: Nodes and Node Types
state and geometry and are described in the section, “pfGeoSet (Geometry Set)” in
Chapter 8. It is important to understand that pfGeoSets are not nodes but are simply
elements of a pfGeode.
Table 3-8 describes the functions for working with pfGeodes.
pfGeode Functions
Table 3-8
Function
Description
pfNewGeode()
Create a pfGeode.
pfAddGSet()
Add a pfGeoSet.
pfRemoveGSet()
Remove a pfGeoSet.
pfInsertGSet()
Insert a pfGeoSet.
pfReplaceGSet()
Replace a pfGeoSet.
pfGetGSet()
Supply a pointer to the specified pfGeoSet.
pfGetNumGSets()
Determine how many pfGeoSets are in the given pfGeode.
Example 3-7 shows how to attach several pfGeoSets to a pfGeode.
Example 3-7
Adding pfGeoSets to a pfGeode
pfGeode *car1;
pfGeoSet *muffler, *frame, *windows, *seats, *tires;
muffler
frame =
seats =
tires =
= read_in_muffler_geometry();
read_in_frame_geometry();
read_in_seat_geometry();
read_in_tire_geometry();
pfAddGSet(car1,
pfAddGSet(car1,
pfAddGSet(car1,
pfAddGSet(car1,
...
72
muffler);
frame);
seats);
tires);
007-1680-100
Node Types
pfText Nodes
A pfText node is a libpf leaf node that contains a set of libpr pfStrings that should be
rendered based on the libpf cull and draw traversals. In this sense, a pfText is similar
to a pfGeode except that it renders 3D text through the libpr pfString and pfFont
mechanisms rather than rendering standard 3D geometry via libpr pfGeoSet and
pfGeoState functionality. pfText nodes are useful for displaying 3D text and other
collections of geometry from a fixed index list. Table 3-9 lists the major pfText functions.
Table 3-9
pfText Functions
Function
Description
pfNewText()
Create a pfText.
pfAddString()
Add a pfString.
pfRemoveString()
Remove a pfString.
pfInsertString()
Insert a pfString.
pfReplaceString()
Replace a pfString.
pfGetString()
Supply a pointer to the specified pfString.
pfGetNumStrings()
Determine how many pfStrings are in the given pfText.
Using the pfText facility is easy. Example 3-8 shows how a pfFont is defined, how
pfStrings are created that reference that font, and then how those pfStrings are added to
a pfText node for display. See the description of pfStrings and pfFonts in Chapter 8,
“Geometry,” for information on setting up individual strings to input into a pfText node.
Example 3-8
Adding pfStrings to a pfText
int nStrings,i;
char tmpBuf[8192];
char fontName[128];
pfFont *fnt = NULL;
/* Create a new text node
pfText *txt = pfNewText();
/* Read in font using libpfdu utility function */
scanf(“%s”,fontName);
fnt = pfdLoadFont(“type1”,fontName,PFDFONT_EXTRUDED);
007-1680-100
73
3: Nodes and Node Types
/* Cant render pfText or libpr pfString without a pfFont */
if (fnt == NULL)
pfNotify(PFNFY_WARN,PFNFY_PRINT,
”No Such Font - %s\n”,fontName);
/* Read nStrings text strings from standard input and */
/* Attach them to a pfText */
scanf(“%d”,&nStrings);
for(i=0;i<nStrings;i++)
{
char c;
int j=0;
int done = 0;
pfString *curStr = NULL;
while(done < 2) /* READ STRING - END on ‘||’ */
{
c = getchar();
if (c == ‘|’)
done++;
else
done = 0;
tmpBuf[j++] = c;
}
tmpBuf[PF_MAX2(j-2,0)] = ‘\0’;
/* Create new libpr pfString structure to attach to pfText */
curStr = pfNewString(pfGetSharedArena());
/* Set the font for the libpr pfString */
pfStringFont(curStr, fnt);
/* Assign the char string to the pfString */
pfStringString(curStr, tmpBuf);
/* Add this libpr pfString to the pfText node */
/* Like adding a libpr pfGeoSet to a pfGeode */
pfAddString(txt, curStr);
}
pfAddChild(SceneGroup, txt);
74
007-1680-100
Node Types
pfBillboard Nodes
A pfBillboard is a pfGeode that rotates its children’s geometry to follow the view
direction or the eyepoint. Billboards are useful for portraying complex objects that are
roughly symmetrical in one or more axes. The billboard rotates to always present the
same image to the viewer using far fewer polygons than a solid model uses. In this way,
billboards reduce both transformation and pixel fill demands on the graphics subsystem
at the expense of some additional host processing. A classic example is a textured
billboard of a single quadrilateral representing a tree.
Because a pfBillboard is also a pfGeode, you can pass a pfBillboard argument to any
pfGeode routine. To add geometry, call pfAddGSet() (see “pfGeode Nodes” on page 71).
Each pfGeoSet in the pfBillboard is treated as a separate piece of billboard geometry; each
one turns so that it always faces the eyepoint.
The pfBillboards can be either constrained to rotate about an axis, as is done for a tree or
a lamp post, or constrained only by a point, as when simulating a cloud or a puff of
smoke. Specify the rotation mode by calling pfBboardMode(); specify the rotational axis
by calling pfBboardAxis(). Since rotating the geometry to the eyepoint does not fully
constrain the orientation of a point-rotating billboard, modes are available to use the
additional degree of freedom to align the billboard in eye space or world space. Usually
the normals of billboards are specified to be parallel to the rotational axis to avoid
lighting anomalies.
The pfFlatten() function is highly recommended for billboards. If a billboard lies beneath
a pfSCS or pfDCS, an additional transformation is done for each billboard. This can have
a substantial performance impact on the cull process, where billboards are transformed.
Table 3-10 describes the functions for working with pfBillboards.
Table 3-10
007-1680-100
pfBillboard Functions
Function
Description
pfNewBboard()
Create a pfBillboard node.
pfBboardPos()
Set a billboard’s position.
pfGetBboardPos()
Find out a billboard’s position.
pfBboardAxis()
Specify the rotation or alignment axis.
pfGetBboardAxis()
Find out the rotation or alignment axis.
75
3: Nodes and Node Types
Table 3-10
pfBillboard Functions (continued)
Function
Description
pfBboardMode()
Specify a billboard’s rotation type.
pfGetBboardMode()
Find out a billboard’s rotation type.
Example 3-9 demonstrates the construction of a pfBillboard node. The code can be found
in /usr/share/Performer/src/pguide/libpf/C/billboard.c for IRIX and
Linux and in %PFROOT%\Src\pguide\libpf\C\billboard.c for Microsoft
Windows.
Example 3-9
Setting Up a pfBillboard
static pfVec2 BBTexCoords[] ={{0.0f,
{1.0f,
{1.0f,
{0.0f,
0.0f},
0.0f},
1.0f},
1.0f}};
static pfVec3 BBVertCoords[4] = /* XZ plane for pt bboards */
{{-0.5f, 0.0f, 0.0f},
{ 0.5f, 0.0f, 0.0f},
{ 0.5f, 0.0f, 1.0f},
{-0.5f, 0.0f, 1.0f}};
static pfVec3 BBAxes[4] = {{1.0f,
{0.0f,
{0.0f,
{0.0f,
0.0f,
1.0f,
0.0f,
0.0f,
0.0f}, /* X */
0.0f}, /* Y */
1.0f}, /* Z */
1.0f}}; /*world Zup*/
static int BBPrimLens[] = { 4 };
static pfVec4 BBColors[] = {{1.0, 1.0, 1.0, 1.0}};
/* Convert static data to pfMalloc’ed data */
static void*
memdup(void *mem, size_t bytes, void *arena)
{
void *data = pfMalloc(bytes, arena);
memcpy(data, mem, bytes);
return data;
}
76
007-1680-100
Node Types
/* For pedagogical use only. Reasonable performance
* requires more then one pfGeoSet per pfBillboard.
*/
pfBillboard*
MakeABill(pfVec3 pos, pfGeoState *gst, long bbType)
{
pfGeoSet *gset;
pfGeoState *gstate;
pfBillboard *bill;
void *arena = pfGetSharedArena();
gset = pfNewGSet(arena);
gstate = pfNewGState(arena);
pfGStateMode(gstate, PFSTATE_ENLIGHTING, PF_OFF);
pfGStateMode(gstate, PFSTATE_ENTEXTURE, PF_ON);
/*.... Create/load texture map for billboard... */
pfGStateAttr(gstate, PFSTATE_TEXTURE, texture);
pfGSetGState(gset, gstate);
pfGSetAttr(gset, PFGS_COORD3, PFGS_PER_VERTEX,
memdup(BBVertCoords, sizeof(BBVertCoords), arena),
NULL);
pfGSetAttr(gset, PFGS_TEXCOORD2, PFGS_PER_VERTEX,
memdup(BBTexCoords, sizeof(BBTexCoords), arena),
NULL);
pfGSetAttr(gset, PFGS_COLOR4, PFGS_OVERALL,
memdup(BBColors, sizeof(BBColors), arena),
NULL);
pfGSetPrimLengths(gset,
(int*)memdup(BBPrimLens, sizeof(BBPrimLens), arena));
pfGSetPrimType(gset, PFGS_QUADS);
pfGSetNumPrims(gset, 1);
pfGSetGState(gset, gst);
bill = pfNewBboard();
switch (bbType)
{
case PF_X: /* axial rotate */
case PF_Y:
case PF_Z:
pfBboardAxis(bill, BBAxes[bbType]);
007-1680-100
77
3: Nodes and Node Types
pfBboardMode(bill, PFBB_ROT, PFBB_AXIAL_ROT);
break;
case 3: /* point rotate */
pfBboardAxis(bill, BBAxes[bbType]);
pfBboardMode(bill, PFBB_ROT, PFBB_POINT_ROT_WORLD);
break;
}
pfAddGSet(bill, gset);
pfBboardPos(bill, 0, pos);
return bill;
}
pfPartition Nodes
A pfPartition is a pfGroup that organizes the scene graphs of its children into a static data
structure that can be more efficient for intersections. Currently, partitions are only useful
for data that lies more or less on an XY plane, such as terrain. Therefore, a pfPartition
would be inappropriate for a skyscraper model.
Partition construction comes in two phases. After a piece of the scene graph has been
placed under the pfPartition, pfBuildPart() examines the spatial arrangement of
geometry beneath the pfPartition and determines an appropriate origin and spacing for
the grid. Because the search is exhaustive, this examination can be time-consuming the
first time through. Once a good partitioning is determined, the search space can be
restricted for future database loads using the partition attributes.
The second phase is invoked by pfUpdatePart(), which distributes the pfGeoSets under
the pfPartition into cells in the spatial partition created by pfBuildPart(). pfUpdatePart()
needs to be called if any geometry under the pfPartition node changes.
During intersection traversal, the segments in a pfSegSet (see “Intersection Requests:
pfSegSets” in Chapter 4) are scan-converted onto the grid, yielding faster access to those
pfGeoSets that potentially intersect the segment. A pfPartition can be made to function
as a normal pfGroup during intersection traversal by ORing PFTRAV_IS_NO_PART into
the intersection traversal mode in the pfSegSet.
78
007-1680-100
Node Types
Table 3-11 describes the functions for working with pfPartitions.
Table 3-11
pfPartition Functions
Function
Description
pfNewPart()
Create a pfPartition.
pfPartVal()
Set the desired pfPartition value.
pfGetPartVal()
Find out the attributes of the specified value.
pfPartAttr()
Set the desired pfPartition attribute.
pfGetPartAttr()
Find out the attributes of specified the attribute.
pfBuildPart()
Construct a spatial partitioning based on the attributes.
pfUpdatePart()
Traverse the partition’s children and incorporate changes.
pfGetPartType()
Determine what kind of partition is being used.
Example 3-10 demonstrates setting up and using a pfPartition node.
Example 3-10
Setting Up a pfPartition
pfGroup *terrain;
pfPartition *partition;
pfScene *scene;
...
terrain = read_in_grid_aligned_terrain();
...
/* create a default partitioning of a terrain grid */
partition = pfNewPart();
pfAddChild(scene, partition);
pfAddChild(partition, terrain);
pfBuildPart(partition);
...
/* use the partitions to perform efficient intersections
* of sets of segments with the terrain */
for(i = 0; i < numVehicles; i++)
007-1680-100
79
3: Nodes and Node Types
pfNodeIsectSegs(partition, vehicle_segment_set[i],
hit_struct);
...
Sample Program
The sample program shown in Example 3-11 demonstrates scene graph construction,
shared instancing, and transformation inheritance. The program uses OpenGL
Performer objects and functions that are described fully in later chapters.
This program reads the names of two objects from the command line, although defaults
are supplied if file names are not given. These files are loaded and a second instance of
each object is created. In each case, this instance is made to orbit the original object, and
the second pair are also placed in orbit around the first. This program is inherit.c and
is part of the suite of OpenGL Performer Programmer’s Guide example programs.
Example 3-11
Inheritance Demonstration Program
/*
* inherit.c - transform inheritance example
*/
#include <math.h>
#include <Performer/pf.h>
#include <Performer/pfdu.h>
int
main(int argc, char *argv[])
{
pfPipe *pipe;
pfPipeWindow *pw;
pfScene *scene;
pfChannel *chan;
pfCoord view;
float z, s, c;
pfNode *model1, *model2;
pfDCS *node1, *node2;
pfDCS *dcs1, *dcs2, *dcs3, *dcs4;
pfSphere sphere;
char *file1, *file2;
/* choose default objects of none specified */
80
007-1680-100
Sample Program
file1 = (argc > 1) ? argv[1] : “blob.nff”;
file2 = (argc > 1) ? argv[1] : “torus.nff”;
/* Initialize Performer */
pfInit();
pfFilePathv(
“.”,
“./data”,
“../data”,
“../../data”,
“../../../data”,
“../../../../data”,
“/usr/share/Performer/data”,
NULL);
/* Single thread for simplicity */
pfMultiprocess(PFMP_DEFAULT);
/* Load all loader DSO’s before pfConfig() forks */
pfdInitConverter(file1);
pfdInitConverter(file2);
/* Configure */
pfConfig();
/* Load the files */
if ((model1 = pfdLoadFile(file1)) == NULL)
{
pfExit();
exit(-1);
}
if ((model2 = pfdLoadFile(file2)) == NULL)
{
pfExit();
exit(-1);
}
/* scale models to unit size */
node1 = pfNewDCS();
pfAddChild(node1, model1);
pfGetNodeBSphere(model1, &sphere);
if (sphere.radius > 0.0f)
pfDCSScale(node1, 1.0f/sphere.radius);
007-1680-100
81
3: Nodes and Node Types
node2 = pfNewDCS();
pfAddChild(node2, model2);
pfGetNodeBSphere(model2, &sphere);
if (sphere.radius > 0.0f)
pfDCSScale(node2, 1.0f/sphere.radius);
/* Create the hierarchy */
dcs4 = pfNewDCS();
pfAddChild(dcs4, node1);
pfDCSScale(dcs4, 0.5f);
dcs3 = pfNewDCS();
pfAddChild(dcs3, node1);
pfAddChild(dcs3, dcs4);
dcs1 = pfNewDCS();
pfAddChild(dcs1, node2);
dcs2 = pfNewDCS();
pfAddChild(dcs2, node2);
pfDCSScale(dcs2, 0.5f);
pfAddChild(dcs1, dcs2);
scene = pfNewScene();
pfAddChild(scene, dcs1);
pfAddChild(scene, dcs3);
pfAddChild(scene, pfNewLSource());
/* Configure and open GL window */
pipe = pfGetPipe(0);
pw = pfNewPWin(pipe);
pfPWinType(pw, PFPWIN_TYPE_X);
pfPWinName(pw, “OpenGL Performer”);
pfPWinOriginSize(pw, 0, 0, 500, 500);
pfOpenPWin(pw);
chan = pfNewChan(pipe);
pfChanScene(chan, scene);
pfSetVec3(view.xyz, 0.0f, 0.0f, 15.0f);
pfSetVec3(view.hpr, 0.0f, -90.0f, 0.0f);
pfChanView(chan, view.xyz, view.hpr);
/* Loop through various transformations of the DCS’s */
for (z = 0.0f; z < 1084; z += 4.0f)
82
007-1680-100
Sample Program
{
pfDCSRot(dcs1,
(z < 360) ? (int) z % 360 : 0.0f,
(z > 360 && z < 720) ? (int) z % 360 : 0.0f,
(z > 720) ? (int) z % 360 : 0.0f);
pfSinCos(z, &s, &c);
pfDCSTrans(dcs2, 1.0f * c, 1.0f * s, 0.0f);
pfDCSRot(dcs3, z, 0, 0);
pfDCSTrans(dcs3, 4.0f * c, 4.0f * s, 4.0f * s);
pfDCSRot(dcs4, 0, 0, z);
pfDCSTrans(dcs4, 1.0f * c, 1.0f * s, 0.0f);
pfFrame();
}
/* show objects static for three seconds */
sleep(3);
pfExit();
exit(0);
}
007-1680-100
83
Chapter 4
4. Database Traversal
Chapter 3, “Nodes and Node Types,” described the node types used by libpf. This
chapter describes the operations that can be performed on the run-time database defined
by a scene graph. These operations typically work with part or all of a scene graph and
are known as traversals because they traverse the database hierarchy. OpenGL Performer
supports four major kinds of database traversals:
•
Application
•
Cull
•
Draw
•
Intersection
The application traversal updates the active elements in the scene graph for the next
frame. This includes processing active nodes and invoking user-supplied callbacks for
animations or other embedded behaviors.
Visual processing consists of two basic traversals: culling and drawing. The cull traversal
selects the visible portions of the database and puts them into a display list. The draw
traversal then runs through that display list and sends rendering commands to the
Geometry Pipeline. Once you have set up all the necessary elements, culling and
drawing are automatic, although you can customize each traversal for special purposes.
The intersection traversal computes the intersection of one or more line segments with
the database. The intersection traversal is user-directed. Intersections are used to
determine the following:
•
Height above terrain
•
Line-of-sight visibility
•
Collisions with database objects
Like other traversals, intersection traversals can be directed by the application through
identification masks and function callbacks. Table 4-1 lists the routines and data types
007-1680-100
85
4: Database Traversal
relevant to each of the major traversals; more information about the listed traversal
attributes can be found later in this chapter and in the appropriate man pages.
Table 4-1
Traversal Attributes for the Major Traversals
Traversal
Attribute
Application
PFTRAV_APP
Cull
PFTRAV_CULL
Draw
PFTRAV_DRAW
Intersection
PFTRAV_ISECT
Controllers
pfChannel
pfChannel
pfChannel
pfSegSet
Global
Activation
pfFrame()
pfSync()
pfAppFrame()
pfFrame()
pfFrame()
pfFrame()
pfNodeIsectSegs(), pfChanNodeIsectSegs()
Global
Callbacks
pfChanTravFunc()
pfChanTravFunc()
pfChanTravFunc()
pfIsectFunc()
Activation
within
Callback
pfApp()
pfCull()
pfDraw()
pfFrame()
pfNodeIsectSegs(), pfChanNodeIsectSegs()
Path
Activation
N/A
pfCullPath()
N/A
N/A
Modes
pfChanTravMode()
pfChanTravMode()
pfChanTravMode()
pfSegSet (also
discriminator
callback)
Node
Callbacks
pfNodeTravFuncs()
pfNodeTravFuncs()
pfNodeTravFuncs()
pfNodeTravFuncs()
Traverser
Masks
pfChanTravMask()
pfChanTravMask()
pfChanTravMask()
pfSegSet mask
Traversee
Masks
pfNodeTravMask()
pfNodeTravMask()
pfNodeTravMask()
pfNodeTravMask()
pfGSetIsectMask()
86
007-1680-100
Scene Graph Hierarchy
Scene Graph Hierarchy
A visual database, also known as a scene, contains state information and geometry. A
scene is organized into a hierarchical structure known as a graph. The graph is composed
of connected database units called nodes. Nodes that are attached below other nodes in
the tree are called children. Children belong to their parent node. Nodes with the same
parent are called siblings.
Database Traversals
The scene hierarchy supplies definitions of how items in the database relate to one
another. It contains information about the logical and spatial organization of the
database. The scene hierarchy is processed by visiting the nodes in depth-first order and
operating on them. The process of visiting, or touching, the nodes is called traversing the
hierarchy. The tree is traversed from top to bottom and from left to right. OpenGL
Performer implements several types of database traversals, including application, clone,
cull, delete, draw, flatten, and intersect. These traversals are described in more detail later
in this chapter.
The principal traversals (application, cull, draw and intersect) all use a similar traversal
mechanism that employs traversal masks and callbacks to control the traversal. When a
node is visited during the traversal, processing is performed in the following order:
1.
Prune the node based on the bitwise AND of the traversal masks of the node and
the pfChannel (or pfSegSet). If pruned, traversal continues with the node’s siblings.
2. Invoke the node’s pre-traversal callback, if any, and either prune, continue, or
terminate the traversal, depending on the callback’s return value.
3. Traverse, beginning again at step 1, the node’s children or geometry (pfGeoSets). If
the node is a pfSwitch, a pfSequence, or a pfLOD, the state of the node affects which
children are traversed.
4. Invoke the node’s post-traversal callback, if any.
State Inheritance
In addition to imposing a logical and spatial ordering of the database, the hierarchy also
defines how state is inherited between parent and child nodes during scene graph
traversals. For example, a parent node that represents a transformation causes the
007-1680-100
87
4: Database Traversal
subsequent transformation of each of its children when it and they are traversed. In other
words, the children inherit state, which includes the current coordinate transformation,
from their parent node during database traversal.
A transformation is a 4x4 homogeneous matrix that defines a 3D transformation of
geometry, which typically consist of scaling, rotation, and translation. The node types
pfSCS and pfDCS both represent transformations. Transformations are inherited through
the scene graph with each new transformation being concatenated onto the ones above
it in the scene graph. This allows chained articulations and complex modeling
hierarchies.
The effects of state are propagated downward only, not from left to right nor upward.
This means that only parents can affect their children—siblings have no effect on each
other nor on their parents. This behavior results in an easy-to-understand hierarchy that
is well suited for high-performance traversals.
Graphics states such as textures and materials are not inherited by way of the scene
graph but are encapsulated in leaf geometry nodes called pfGeode nodes, which are
described in the section “Node Types” in Chapter 3.
Database Organization
OpenGL Performer uses the spatial organization of the database to increase the
performance of certain operations such as drawing and intersections. It is therefore
recommended that you consider the spatial arrangement of your database. What you
might think of as a logical arrangement of items in the database may not match the
spatial arrangement of those items in the visual environment, which can reduce OpenGL
Performer’s ability to optimize operations on the database. See “Organizing a Database
for Efficient Culling” on page 94 for more information about spatial organization in a
visual database and the efficiency of database operations.
Application Traversal
The application traversal is the first traversal that occurs during the processing of the
scene graph in preparation for rendering a frame. It is initiated by calling pfAppFrame().
If pfAppFrame() is not explicitly called, the traversal is automatically invoked by
pfSync() or pfFrame(). An application traversal can be invoked for each channel, but
usually channels share the same application traversal (see pfChanShare()).
88
007-1680-100
Application Traversal
The application traversal updates dynamic elements in the scene graph, such as
geometric morphing. The application traversal is also often used for implementing
animations or other custom processing when it is desirable to have those behaviors
embedded in the scene graph and invoked by OpenGL Performer rather than requiring
application code to invoke them every frame.
The traversal proceeds as described in “Database Traversals.” The selection of which
children to traverse is also affected by the application traversal mode of the channel, in
particular the choice of all, none, or one of the children of pfLOD, pfSequence and
pfSwitch nodes is possible. By default, the traversal obeys the current selection dictated
by these nodes.
The following example (this loader reads both Open Inventor and VRML files) shows a
simple callback changing the transformation on a pfDCS every frame.
Example 4-1
Application Callback to Make a Pendulum
int
AttachPendulum(pfDCS *dcs, PendulumData *pd)
{
pfNodeTravFuncs(dcs, PFTRAV_APP, PendulumFunc, NULL);
pfNodeTravData(dcs, PFTRAV_APP, pd);
}
int
PendulumFunc(pfTraverser *trav, void *userData)
{
PendulumData *pd = (PendulumData*)userData;
pfDCS *dcs = (pfDCS*)pfGetTravNode(trav);
if (pd->on)
{
pfMatrix mat;
double now = pfGetFrameTimeStamp();
float frac, dummy;
pd->lastAngle += (now - pd->lastTime)*360.0f*pd->frequency;
if (pd->lastAngle > 360.0f)
pd->lastAngle -= 360.0f;
// using sinusoidally generated angle
pfSinCos(pd->lastAngle, &frac, &dummy);
frac = 0.5f + 0.5f * frac;
frac = (1.0f - frac)*pd->angle0 + frac*pd->angle1;
007-1680-100
89
4: Database Traversal
pfMakeRotMat(mat,
frac, pd->axis[0], pd->axis[1], pd->axis[2]);
pfDCSMat(dcs, mat);
pd->lastTime = now;
}
return PFTRAV_CONT;
}
Cull Traversal
The cull traversal occurs in the cull phase of the libpf rendering pipeline and is initiated
by calling pfFrame(). A cull traversal is performed for each pfChannel and determines
the portion of the scene to be rendered. The traversal processes the subgraphs of the
scene that are both visible and selected by nodes in the scene graph that control traversal
(that is, pfLOD, pfSequence, pfSwitch). The visibility culling itself is performed by
testing bounding volumes in the scene graph against the channel’s viewing frustum.
For customizing the cull traversal, libpf provides traversal masks and function
callbacks for each node in the database, as well as a function callback in which the
application can do its own culling of custom data structures.
Traversal Order
The cull is a depth-first, left-to-right traversal of the database hierarchy beginning at a
pfScene, which is the hierarchy’s root node. For each node, a series of tests is made to
determine whether the traversal should prune the node—that is, eliminate it from further
consideration—or continue on to that node’s children. The cull traversal processing is
much as described earlier; in particular, the draw traversal masks are compared and the
node is checked for visibility before the traversal continues on to the node’s children.
Processing proceeds in the following order:
1.
Prune the node, based on the channel’s draw traversal mask and the node’s draw
mask.
2. Invoke the node’s pre-cull callback and either prune, continue, or terminate the
traversal, depending on callback’s return value.
3. Prune the node if its bounding volume is completely outside the viewing frustum.
90
007-1680-100
Cull Traversal
4. Traverse, beginning again at step 1, the node’s children or geometry (pfGeoSets) if
the node is completely or partially in the viewing frustum. If the node is a pfSwitch,
a pfSequence, or a pfLOD, the state of the node affects which children are traversed.
5. Invoke the node’s post-cull callback.
The following sections discuss these steps in more detail.
Visibility Culling
Culling determines whether a node is within a pfChannel’s viewing frustum for the
current frame. Nodes that are not visible are pruned—omitted from the list of objects to
be drawn—so that the Geometry Pipeline does not waste time processing primitives that
couldn’t possibly appear in the final image.
Hierarchical Bounding Volumes
Testing a node for visibility compares the bounding volume of each object in the scene
against a viewing frustum that is bounded by the near and far clip planes and the four
sides of the viewing pyramid. Both nodes (see Chapter 3, “Nodes and Node Types”) and
pfGeoSets (see Chapter 8, “Geometry”) have bounding volumes that surround the
geometry that they contain. Bounding volumes are simple geometric shapes whose
centers and edges are easy to locate. Bounding volumes are organized hierarchically so
that the bounding volume of a parent encloses the bounding volumes of all its children.
You can specify bounding volumes or let OpenGL Performer generate them for you (see
“Bounding Volumes” in Chapter 3).
Figure 4-1 shows a frustum and three objects surrounded by bounding boxes. Two of the
objects are outside the frustum; one is within it. One of the objects outside the frustum
has a bounding box whose edges intersect the frustum, as shown by the shaded area. The
visibility test for this object returns TRUE, because its bounding box does intersect the
view frustum even though the object itself does not.
007-1680-100
91
4: Database Traversal
PFIS_FALSE
PFIS_ALL_IN
PFIS_TRUE
Figure 4-1
92
Culling to the Frustum
007-1680-100
Cull Traversal
Visibility Testing
The cull traversal begins at the root node of a channel’s scene graph (the pfScene node)
and continues downward, directed by the results of the cull test at each node. At each
node the cull test determines the relationship of the node’s bounding volume to the
viewing frustum. Possible results are that the bounding volume is entirely outside, is
entirely within, is partially within, or completely contains the viewing frustum.
If the intersection test indicates that the bounding volume is entirely outside the frustum,
the traversal prunes that node—that is, it does not consider the children of that node and
continues with the node’s next sibling.
If the intersection test indicates that the bounding volume is entirely inside the frustum,
the node’s children are not cull-tested because the hierarchical nature of bounding
volumes implies that the children must also be entirely within the frustum.
If the intersection test indicates that the bounding volume is partially within the frustum,
or that the bounding volume completely contains the frustum, the testing process
continues with the children of that node. Because a bounding volume is larger than the
object it surrounds, it is possible for a bounding volume to be partially within a frustum
even when none of its enclosed geometry is visible.
By default, OpenGL Performer tests bounding volumes all the way down to the pfGeoSet
level (see Chapter 8, “Geometry”) to provide fine-grained culling. However, if your
application is spending too much time culling, you can stop culling at the pfGeode level
by calling pfChanTravMode(). Then if part of a pfGeode is potentially visible, all
geometry in that pfGeode is drawn without cull-testing it first.
Visibility Culling Example
Figure 4-2 portrays a simple database that contains a toy block, train, and car. The block
is outside the frustum, the bounding volume of the car is partially inside the frustum, and
the train is completely inside the frustum.
007-1680-100
93
4: Database Traversal
Figure 4-2
Sample Database Objects and Bounding Volumes
Organizing a Database for Efficient Culling
Efficient culling depends on having a database whose hierarchy is organized spatially. A
good technique is to partition the database into regions, called tiles. Tiling is also required
for database paging. Instead of culling the entire database, only the tiles that are within the
view frustum need to be traversed.
94
007-1680-100
Cull Traversal
The worst case for the cull traversal performance is to have a very flat hierarchy—that is,
a pfScene with all the pfGeodes directly under it and many pfGeoSets in each pfGeode—
or a hierarchy that is organized by object type (for example, having all trees in the
database grouped under one pine tree node, rather than arranged spatially).
Figure 4-3 shows a sample database represented by cubes, cones, pyramids, and spheres.
Organizing this database spatially, rather than by object type, promotes efficient culling.
This type of spatial organization is the most effective control you have over efficient
traversal.
007-1680-100
95
4: Database Traversal
Board
Pyramids
Cones
Spheres
Cubes
Board
Tile 1
Figure 4-3
96
Tile 2
Tile 3
Tile 4
Tile 5
Tile 6
Tile 7
Tile 8
Tile 9
How to Partition a Database for Maximum Efficiency
007-1680-100
Cull Traversal
When modeling a database, you should consider other trade-offs as well. Small amounts
of geometry in each culling unit, whether pfGeode or pfGeoSet, provide better culling
resolution and result in sending less nonvisible geometry to the pipeline. Small pieces
also improve the performance of line-segment intersection inquiries (see “Database
Concerns” in Chapter 24). However, using many small pieces of geometry can increase
the traversal time and can also reduce drawing performance. The optimal amount of
geometry to place in each pfGeoSet depends on the application, database, system CPU,
and graphics hardware.
Custom Visibility Culling
Existence within the frustum is not the only criterion that determines an object’s
visibility. The item may be too distant to be seen from the viewpoint, or it may be
obscured by other objects between it and the viewer, such as a wall or a hill. Atmospheric
conditions can also affect object visibility. An object that is normally visible at a certain
distance may not be visible at that same distance in dense fog.
Implementing more sophisticated culling requires knowledge of visibility conditions
and control over the cull traversal. The cull traversal can be controlled through traversal
masks, which are described in the section titled “Controlling and Customizing
Traversals” on page 102.
Knowing whether an object is visible requires either prior information about the spatial
organization of a database, such as cell-to-cell visibilities, or run-time testing such as
computing line-of-sight visibility (LOS). You can compute simple LOS visibility by
intersecting line segments that start at the eyepoint with the database. See the
“Intersection Traversal” on page 120.
Sorting the Scene
During the cull traversal, a pfChannel can rearrange the order in which pfGeoSets are
rendered for improved performance and image quality. It does this by binning and
sorting. Binning is the act of placing pfGeoSets into specific bins, which are rendered in a
specific order. OpenGL Performer provides two default bins: one for opaque geometry
and one for blended, transparent geometry. The opaque bin is drawn before the
transparent bin so transparent surfaces are properly blended with the background scene.
Applications are free to add new bins and specify arbitrary bin orderings.
Bins are often used to group geometry with certain desired characteristics. Sometimes it
may be desirable for a pfGeoSet to be in several bins. For this purpose you can create a
007-1680-100
97
4: Database Traversal
sub-bin of two existing bins using the function pfChanFindSubBin(bin1,bin2,int). The
parameters are the two parent bins and an integer value indicating whether the sub-bin
should be created if it does not exist. The function returns a value of –1 if the bin does not
exist (and it was not supposed to be created) or if any of the parent bins do not exist. If
you need to create a sub-bin of more than two bins, call this function several times. For
example, to create a sub-bin of bin 5, 6, and 7, you call pfChanFindSubBin() with
parameters 5 and 6. Let us assume that sub-bin of bin 5 and 6 is bin 8. Then you call
pfChanFindSubBin() again with parameters 8 and 7 to obtain a sub-bin of bins 5, 6, and
7. It does not matter in what order you call it because all sub-bins are directly linked to
their parent root bins (and vice versa); there is no tree hierarchy. See the section “Cull
Programs” on page 103 for an example of using sub-bins.
The function pfChanFindBinParent(bin,int) returns the first parent of bin bin that is
bigger than the value specified as the second parameter. Thus, by calling this method
several times (until it returns –1), you can determine all parents of a bin.
Sorting is done on a per-bin basis. pfGeoSets within a given bin are sorted by a specific
criterion. Two useful criteria provided by OpenGL Performer are sorting by graphics
state and sorting by range. When sorting by state, pfGeoSets are sorted first by their
pfGeoState, then by an application-specified hierarchy of state modes, values, and
attributes which are identified by PFSTATE_* tokens and are described in Chapter 12,
“Graphics State”. State sorting can offer a huge performance advantage since it greatly
reduces the number of mode changes carried out by the Geometry Pipeline. State sorting
is the default sorting configuration for the opaque bin. If a bin has sub-bins, pfGeoSets
are ordered in each sub-bin separately, as are pfGeoSets that do not belong to any sub-bin
of the bin.
Range sorting is required for proper rendering of blended, transparent surfaces, which
must be rendered in back-to-front order so that each surface is properly blended with the
current background color. Front-to-back sorting is also supported. The default sorting for
the transparent bin is back-to-front sorting. Note that the sorting granularity is
per-pfGeoSet, not per-triangle so transparency sorting is not perfect.
In case of the transparent bin, the order in which pfGeoSets are drawn (back-to-front) is
important to avoid visible artifacts. Sub-bins, even if their pfGeoSets were ordered
back-to-front, may break that order. For this purpose, you can mark selected bins as
non-exclusive. If a pfGeoSet belongs to a sub-bin of a non-exclusive bin, it is added both
to the sub-bin and directly to the list of pfGeoSets of the non-exclusive bin. Thus, when
pfGeoSets of a non-exclusive bin are sorted, they are all in one list. Any root bin can be
marked non-exclusive by setting flag PFBIN_NONEXCLUSIVE_BIN using the function
pfChanBinFlags(). The transparent bin is by default non-exclusive.
98
007-1680-100
Cull Traversal
The pfChannel bins are given a rendering order and a sorting configuration with
pfChanBinOrder() and pfChanBinSort(), respectively. A sub-bin inherits sorting
configuration from a parent with the highest sort priority, set by
pfChanBinSortPriority(). Sorting configuration of sub-bins cannot be changed using
pfChanBinOrder().
A bin’s order is simply an integer identifying its place in the list of bins. An order less
than 0 or PFSORT_NO_ORDER means that pfGeoSets that fall into the bin are drawn
immediately without any ordering or sorting. Multiple bins may have the same order but
the rendering precedence among these bins is undefined. The rendering order of
sub-bins is determined by the child-order mask of their parents. This mask can be set by
pfChanBinChildOrderMask(). When a sub-bin is created, the mask of all its parents is
combined (using a binary OR) and set as a rendering order of the sub-bin.
A bin’s sorting configuration is given as a token identifying the major sorting criterion
and then an optional list of tokens, terminated with the PFSORT_END token, that defines
a state sorting hierarchy. The following tokens control the sort:
PFSORT_BY_STATE
pfGeoSets are sorted first by pfGeoState then by the state elements
found between the PFSORT_STATE_BGN and PFSORT_STATE_END
tokens, for example.
PFSORT_FRONT_TO_BACK
pfGeoSets are sorted by nearest to farthest range from the eyepoint.
Range is computed from eyepoint to the center of the pfGeoSet’s
bounding volume.
PFSORT_BACK_TO_FRONT
pfGeoSets are sorted by farthest to nearest range from the eyepoint.
Range is computed from eyepoint to the center of the pfGeoSet’s
bounding volume.
PFSORT_QUICK
A special, low-cost sorting technique. pfGeoSets must fall into a bin
whose order is 0 in which case they will be sorted by pfGeoState and
drawn immediately. This is the default sorting mode for the
PFSORT_OPAQUE_BIN bin.
For example, the following specification will sort the opaque bin by pfGeoState, then by
pfTexture, then by pfMaterial:
007-1680-100
99
4: Database Traversal
static int sort[] = {PFSORT_STATE_BGN,
PFSTATE_TEXTURE, PFSTATE_FRONTMTL,
PFSORT_STATE_END, PFSORT_END};
pfChanBinSort(chan, PFSORT_OPAQUE_BIN, PFSORT_BY_STATE,
sort);
A pfGeoSet’s draw bin may be set directly by the application with pfGSetDrawBin().
Otherwise, OpenGL Performer automatically determines if the pfGeoSet belongs in the
default opaque or transparent bins. Based on the position of a pfGeoSet in the scene, cull
programs, if they are enabled, can determine the draw bin of the pfGeoSet. See section
“Cull Programs” on page 103 for more details.
Paths through the Scene Graph
You can define a chain, or path, of nodes in a scene graph using the pfPath data structure.
(Note that a pfPath has nothing to do with filesystem paths as specified with the PFPATH
environment variable or with specifying a path for a user to travel through a scene.) Once
you have specified a pfPath with a call to pfNewPath(), you can traverse and cull that
path as a subset of the entire scene graph using pfCullPath(). The function pfCullPath()
must only be called from the cull callback function set by pfChanTravFunc()—see
“Process Callbacks” on page 116 for details. For more information about the pfPath
structure, see the pfPath(3pf) and pfList(3pf) man pages.
When OpenGL Performer looks for intersections, it can return a pfPath to the node
containing the intersection. This feature is particularly useful when you are using
instancing, in which case you cannot use pfGetParent() to find out where in the scene
graph the given node is. Finding out the pfPath to a given node is also useful in
implementing picking.
Draw Traversal
For each bin the cull traversal generates a libpr display list of geometry and state
commands (see “Display Lists” in Chapter 12), which describes the bin's geometry that
is visible from a pfChannel. The draw traversal parses all root bins (bins without a parent
bin) in the order given by their rendering order value. For each root bin, it simply
traverses the display list and sends commands to the Geometry Pipeline to generate the
image. If a bin has sub-bins, objects that are not in any sub-bin of the bin are rendered
100
007-1680-100
Draw Traversal
first and are followed by objects of each sub-bin. The order in which sub-bins of the bin
are drawn is determined by their rendering order value.
Optimizing the Drawing of Sub-bins
To avoid drawing sub-bins multiple times (for each of its parents), set the flag
PFBIN_DONT_DRAW_BY_DEFAULT for those root bins that share sub-bins with the
default opaque or transparent bin. The bin flags can be set using the function
pfChanBinFlags().
Individual bins, including sub-bins, may be rendered with the function pfDrawBin()
called from the draw callback of pfChannel (see pfChanTravFunc()). The bin –1 is a
special pfDrawBin() argument that lets you render the default scene display list, which
contains all the objects that did not fall in any defined bin. Note that this default scene
display list exists only in PFMP_CULL_DL_DRAW multiprocessing mode. In the case of
drawing a sub-bin, all sub-bins that have the same parents as a given sub-bin will be
drawn. For example, consider root bins 5, 6, and 7 and sub-bins 8 (child of 5 and 6) and
9 (child of 5, 6, and 7). When pfDrawBin() is called with bin 8, bin 9 will be rendered as
well.
Traversing a pfDispList is much faster than traversing the database hierarchy because the
pfDispList flattens the hierarchy into a simple, efficient structure. In this way, the cull
traversal removes much of the processing burden from the draw traversal; throughput
greatly increases when both traversals are running in parallel.
Bin Draw Callbacks
Root bins can have draw callbacks associated with them. Draw callbacks are set by
calling function pfChanBinCallBack(). The parameters are the bin number, the type of a
callback (PFBIN_CALLBACK_PRE_DRAW or PFBIN_CALLBACK_POST_DRAW), and
the callback itself. The callback is a function that has only one parameter, a void pointer
that points to the user data. Each bin has one user-data pointer, shared between pre-draw
and post-draw callbacks. This pointer can be set using function pfChanBinUserData().
If the callbacks are costly, it makes sense to group sub-bins of a bin with costly callbacks
together. To achieve this, ensure that you set a high child-order mask for the bin (see the
section “Sorting the Scene” on page 97).
007-1680-100
101
4: Database Traversal
Controlling and Customizing Traversals
The result of the cull traversal is a display list of geometry to be rendered by the draw
traversal. What gets placed in the display list is determined by both visibility and by
other user-specified modes and tests.
pfChannel Traversal Modes
The PFTRAV_CULL argument to pfChanTravMode() modifies the culling traversal. The
cull mode is a bitmask that specifies the modes to enable; it is formed by the logical OR
of one or more of these tokens:
•
PFCULL_VIEW
•
PFCULL_GSET
•
PFCULL_SORT
•
PFCULL_IGNORE_LSOURCES
•
PFCULL_PROGRAM
Culling to the view frustum is enabled by the PFCULL_VIEW token. Culling to the
pfGeoSet-level is enabled by the PFCULL_GSET token and can produce a tighter cull that
improves rendering performance at the expense of culling time.
PFCULL_SORT causes the cull to sort geometry by state—for example, by texture or by
material, in order to optimize rendering performance. It also causes transparent
geometry to be drawn after opaque geometry for proper transparency effects.
By default, the enabled culling modes are PFCULL_VIEW | PFCULL_GSET |
PFCULL_SORT. It is recommended that these modes be enabled unless the cull traversal
becomes a significant bottleneck in the processing pipeline. In this case, try disabling
PFCULL_GSET first, then PFCULL_SORT.
Normally, a pfChannel’s cull traversal pre-traverses the scene, following all paths from
the scene to all pfLightSources in the scene so that light sources can be set up before the
normal scene traversal. If you want to disable this pre-traversal, set the
PFCULL_IGNORE_LSOURCES cull-enable bit but your pfLightSources will not
illuminate the scene.
102
007-1680-100
Controlling and Customizing Traversals
When the PFCULL_PROGRAM token is set, a cull program attached to the channel is
executed for each pfGeoSet during the cull traversal. See the following section “Cull
Programs” for more details.
The PFTRAV_DRAW argument to pfChanTravMode() modifies the draw traversal. A
mode of PFDRAW_ON is the default and will cause the pfChannel to be rendered. A
mode of PFDRAW_OFF indicates that the pfChannel should not be drawn and
essentially turns off the pfChannel.
Cull Programs
This section uses the following topics to describe cull programs:
•
“The pfCullProgram Class” on page 103
•
“Polytopes” on page 104
•
“Predefined Cull Program Instructions” on page 105
•
“User-Defined Cull Program Instructions” on page 109
•
“Cull Traversal” on page 110
•
“Occlusion Culling Using Cull Programs” on page 110
•
“Small-Feature Culling Using Cull Programs” on page 112
The pfCullProgram Class
A pfCullProgram is a class that is used to set up a cull program, a sequence of instructions
that are executed for each scene graph node and each pfGeoSet during the cull traversal.
There can be two separate cull programs, one for scene graph nodes and one for
pfGeosets. The node cull program uses a set of polytopes. Based on the position of the
node with respect to each polytope (inside, outside, intersects), it can determine whether
the node is culled out (good for occlusion culling) or whether all pfGeoSets under this
node are assigned to a specific bin. The pfGeoSet cull program also uses a set of polytopes
and assigns each pfGeoSet to a different bin based on the position of the pfGeoSet with
respect to each polytope or it culls out the pfGeoSet. The best use of cull programs is for
occlusion culling (see section “Occlusion Culling Using Cull Programs” on page 110) or
in multipass rendering when in some passes only parts of the scene have to be rendered.
Being able to assign these parts to a bin can reduce the rendering time.
007-1680-100
103
4: Database Traversal
There is always a default cull program present on a pfChannel. To access it, you can call
pfGetChanCullProgram(). Then you can set the program's instructions and the
polytopes and can enable the cull program by setting token PFCULL_PROGRAM using
the function pfChanTravMode(). See the previous section “pfChannel Traversal Modes”
for more details.
The following code sequence illustrates the use of a cull program:
pfCullProgram *cullPgm = pfGetChanCullProgram(channel);
pfCullProgramResetPgm(cullPgm, PFCULLPG_GEOSET_PROGRAM);
pfCullProgramAddPgmOpcode(cullPgm, opcode1, data1);
...
pfCullProgramAddPgmOpcode(cullPgm, opcodeN, dataN);
pfCullProgramResetPgm(cullPgm, PFCULLPG_NODE_PROGRAM);
pfCullProgramAddPgmOpcode(cullPgm, opcode1, data1);
...
pfCullProgramAddPgmOpcode(cullPgm, opcodeM, dataM);
pfCullProgramNumPolytopes(cullPgm, 2);
pfCullProgramPolytope(cullPgm, ptope1);
pfCullProgramPolytope(cullPgm, ptope2);
You can define both a node and a pfGeoSet cull program at once by setting the token in
pfCullProgramResetPgm() to PFCULLPG_GEOSET_PROGRAM |
PFCULLPG_NODE_PROGRAM.
Polytopes
Cull program polytopes are standard pfPolytopes. They can be used to define various
areas: the area could be some subset of a view frustum in which the geometry is drawn
using special attributes, it could be a bounding box around some area of interest, and so
on.
To initialize cull program polytopes, you set the number of polytopes that are used by the
cull programs using pfCullProgramNumPolytopes(). Then create a new pfPolytope in
the shared arena and set it using pfCullProgramPolytope(). The polytopes are indexed
from 0. Polytopes are shared between the node and the pfGeoset cull program even if the
programs are different.
104
007-1680-100
Controlling and Customizing Traversals
To modify a polytope of a certain index during the simulation, you get a pointer to the
polytope using pfGetCullProgramPolytope(), modify it, and then call
pfCullProgramPolytope().
Predefined Cull Program Instructions
A cull program is a set of instructions that operate on bins and predefined polytopes. You
define instructions one-by-one in the order of their desired execution. First, you use the
function pfCullProgramResetPgm() to reset the default program, which consists of a
return instruction only. Then you specify each instruction by its opcode (predefined
instruction specified with the function pfCullProgramAddPgmOpcode()) or directly by
specifying a user-defined instruction using pfCullProgramAddPgmInstruction(). For
the details of user-defined instructions, see the later section “User-Defined Cull Program
Instructions” on page 109. Each instruction has an associated integer value that is used
as a parameter for the instruction.
The cull program starts with the bin that is associated with the pfGeoSet. As the cull
program executes, it modifies the pfGeoSet. The output is a new bin assignment.
The following categories of predefined instructions are available:
007-1680-100
•
Test instructions
•
Assign instructions
•
Add-bin instructions
•
Jump instructions
•
Return instruction
105
4: Database Traversal
Test Instructions
Table 4-2 describes the test instructions.
Table 4-2
Test Instructions
Instruction
Description
PFCULLPG_TEST_POLYTOPE n
Tests the bounding box of the pfGeoSet or the
bounding sphere of the node against the polytope
with index n. The result is one of PFIS_FALSE (all
out), PFIS_MAYBE (possible intersection), or
PFIS_MAYBE | PFIS_TRUE (all in).
PFCULLPG_TEST_IS_SUBBIN_OF b
Tests whether the bin that has been determined up to
this point is a sub-bin of bin b. The result is 1 or 0.
Note that bin b is considered a sub-bin of itself.
PFCULLPG_TEST_IS_TRANSPARENT
Tests whether the pfGeoSet is transparent. The
parameter is ignored. The result is 1 or 0.
PFCULLPG_TEST_IS_LIGHT_POINT
Tests whether the pfGeoSet belongs to a light point
bin. The parameter is ignored. The result is 1 or 0.
Assign Instructions
Table 4-3 describes the assign instructions.
Table 4-3
106
Assign Instructions
Instruction
Description
PFCULLPG_ASSIGN_BIN_MAYBE b
Assigns bin b to the pfGeoSet if the result of the last
polytope test was PFIS_MAYBE.
PFCULLPG_ASSIGN_BIN_TRUE b
Assigns bin b to the pfGeoSet if the result of the last
binary test was 1.
PFCULLPG_ASSIGN_BIN_ALL_IN b
Assigns bin b to the pfGeoSet if the result of the last
polytope test was PFIS_MAYBE | PFIS_TRUE.
PFCULLPG_ASSIGN_BIN_ALL_OUT b
Assigns bin b to the pfGeoSet if the result of the last
polytope test was PFIS_FALSE.
007-1680-100
Controlling and Customizing Traversals
Table 4-3
Assign Instructions (continued)
Instruction
Description
PFCULLPG_ASSIGN_BIN_FALSE b
Assigns bin b to the pfGeoSet if the result of the last
binary test was 0.
PFCULLPG_ASSIGN_BIN b
Assigns bin b to the pfGeoSet.
Add-Bin Instructions
For each PFCULLPG_ASSIGN* instruction, there is an equivalent
PFCULLPG_ADD_BIN* instruction, in which the pfGeoSet is assigned a sub-bin of bin b
and the existing bin. If the existing bin is –1, the instruction operates as an assign
instruction. If the sub-bin does not exist, it is dynamically created.
The following are the add-bin instructions:
•
PFCULLPG_ADD_BIN_MAYBE b
•
PFCULLPG_ADD_BIN_TRUE b
•
PFCULLPG_ADD_BIN_ALL_IN b
•
PFCULLPG_ADD_BIN_ALL_OUT b
•
PFCULLPG_ADD_BIN_FALSE b
•
PFCULLPG_ADD_BIN b
Jump Instructions
Table 4-4 describes the jump instructions.
Table 4-4
007-1680-100
Jump Instructions
Instruction
Description
PFCULLPG_JUMP_MAYBE c
Skips next c instructions if the result of the last polytope
test was PFIS_MAYBE. If c is negative, go back |c|–1
instructions.
PFCULLPG_JUMP_TRUE c
Skips next c instructions if the result of the last binary test
was 1.
PFCULLPG_JUMP_ALL_IN c
Skips next c instructions if the result of the last polytope
test was PFIS_MAYBE | PFIS_TRUE.
107
4: Database Traversal
Table 4-4
Jump Instructions (continued)
Instruction
Description
PFCULLPG_JUMP_ALL_OUT c
Skips next c instructions if the result of the last polytope
test was PFIS_FALSE.
PFCULLPG_JUMP_FALSE c
Skips next c instructions if the result of the last polytope
test was 0.
PFCULLPG_JUMP c
Skips next c instructions.
Return Instruction
Each cull program must be terminated by a return instruction. You specify the return
instruction with PFCULLPG_RETURN flags. The PFCULLPG_RETURN parameter is a
combination of the following binary flags:
•
PFCULLPG_CULL
•
PFCULLPG_CULL_ON_ALL_IN
•
PFCULLPG_CULL_ON_ALL_OUT
•
PFCULLPG_DONT_TEST_TRANSPARENCY
•
PFCULLPG_TEST_LPOINTS
•
PFCULLPG_DONT_TEST_LPOINTS
The first three flags determine whether the node or the pfGeoSet is culled out, optionally
based on the result of the last polytope test. In that case, any bin assignment made by the
cull program is ignored.
The last three flags control whether an additional test for the pfGeoSet being transparent
or being a light point is performed. These flags affect only the pfGeoset cull program. If
the pfGeoSet is transparent or is a light point, the pfGeoSet is assigned the bin resulting
from the cull program and either a sub-bin of the transparent bin or the light point bin.
If initially the pfGeoSet has no bin assigned to it, both the transparency and light point
tests are performed (to follow the operation of a regular cull traversal). If those tests are
not needed, you can use the two DONT_TEST flags. If the pfGeoSet has initially assigned
a bin, the tests are not performed unless the binary flags specify so.
108
007-1680-100
Controlling and Customizing Traversals
If you need to perform any of these two tests earlier—for example, to differentiate bin
assignment based on transparency—you can use the instructions
PFCULLPG_TEST_IS_TRANSPARENT and PFCULLPG_TEST_IS_LIGHT_POINT.
User-Defined Cull Program Instructions
You may provide your own cull program instructions. Each instruction must be a
function that takes two parameters: a pointer to pfCullProgram and an integer value (the
instruction parameter). The instruction has to return a value by which the instruction
counter is increased. This value is 1 for all instructions, except jump instructions.
Actually, it is possible to write whole cull programs as a single, user-defined instruction.
There are two variables that a user-defined instruction can access during the execution
of a cull program and there are several useful methods they may use. The following are
the two variables:
currentResult
Result of the last polytope test or a binary test
bbox
Bounding box for the pfGeoSet
Table 4-5 describes the four functions that can be used in instructions.
Table 4-5
007-1680-100
Functions Available for User-Defined Cull Program Instructions
Function
Description
pfCullProgramTestPolytope(pgm,n)
Tests polytope n using bbox, the bounding box of the
pfGeoSet or the bounding sphere of the node. Use
this function rather than doing the test directly
because the result is often already known by testing
the nodes above the current pfGeoSet or the current
node and the test can be avoided.
pfCullProgramAddBinParent(pgm,b)
Adds a new parent b, which could also be a sub-bin.
The cull program keeps the list of parents that
identify the current bin (to avoid creating many
sub-bins that may not be needed). This function adds
a new parent b, which can be a sub-bin also.
pfCullProgramIsSubbinOf(pgm,b)
Tests whether the current bin is a sub-bin of bin b.
pfCullProgramResetBinParents(pgm)
Resets the list of parents of the current bin.
109
4: Database Traversal
For example, the predefined instruction PFCULLPG_ASSIGN_BIN_MAYBE can be
implemented as shown in the following code:
int MyAssignBinMaybe(pfCullProgram *pgm, int data)
{
if(pgm->currentResult & PFIS_MAYBE)
{
pfCullProgramResetBinParents(pgm);
pfCullProgramAddBinParent(pgm, data);
}
return 1;
}
Cull Traversal
To reduce the amount of testing performed for each pfGeoSet, each node of the tree is
tested against all cull program polytopes when cull programs are enabled. If the test is
conclusive—that is, the bounding sphere of the node is inside or outside of a polytope—
children of the node are not tested against the given polytope. Use the node cull program
to determine culling and use the pfGeoset cull program to assign bins at the pfGeoset
level.
If culling to the view frustum is enabled (token PFCULL_VIEW set by
pfChanTravMode()), it is done before the cull program is executed. In this case, nodes
and pfGeoSets that are not intersecting the view frustum are culled out and the cull
program is not executed for them.
Sample code illustrating the use of cull programs can be found in the following files in
the directory /usr/share/Performer/src/pguide/libpf/C++ on IRIX and Linux
and in %PFROOT%\Src\pguide\libpf\C++ on Microsoft Windows:
•
cullPgmSimple
•
cullPgmMultipass
Occlusion Culling Using Cull Programs
In order to use cull programs for occlusion culling you must choose the occluders in the
scene—for example, walls in a room or the walls of the nearest buldings in a city. Then
you must create a polytope around each occluder. If the occluder is a rectangle, the
polytope must have one face for the rectangle and four faces for the edges, four planes
each defined by the edge and the eye. You must update the polytope or polytopes every
time the eye or the occluder moves.
110
007-1680-100
Controlling and Customizing Traversals
For an example of occlusion culling, see the following program:
/usr/share/Performer/src/pguide/libpf/C++/occlusionCull.C
(IRIX and Linux)
%PFROOT%\Src\pguide\libpf\C++\occlusionCull.cxx
(Microsoft Windows)
The sample program also includes a function that creates a polytope for a polygon
defined by four vertices.
At present the polytopes must be convex. Consequently, in the case that two occluders
are touching but their common shape is concave, they must be defined as two polytopes.
In this case, the geometry that is occluded by both occluders and that spans their
common boundary is not culled out. To avoid this problem, you can define a convex
polytope that contains both shapes (a convex hull) and then define convex cut areas that
are not part of the occluders. In this way, you can also add holes in occluders. As long as
you start with a convex polytope, you can subtract as many convex polytopes as you
need.
The following code sequence illustrates the use of a convex hull and cut-out areas:
PFCULLPG_TEST_POLYTOPE, 0 // convex hull
PFCULLPG_JUMP_ALL_IN, 1
PFCULLPG_RETURN, 0
// no cull
PFCULLPG_TEST_POLYTOPE, 1 // first cutout area
PFCULLPG_JUMP_ALL_OUT, 1
PFCULLPG_RETURN, 0
// no cull
...
PFCULLPG_TEST_POLYTOPE, N // n-th cutout area
PFCULLPG_JUMP_ALL_OUT, 1
PFCULLPG_RETURN, 0
// no cull
PFCULLPG_RETURN, PFCULLPG_CULL
For a complete example, see the following program:
/usr/share/Performer/src/pguide/libpf/C++/occlusionCullConcave.C
(IRIX and Linux)
%PFROOT%\Src\pguide\libpf\C++\occlusionCullConcave.cxx
(Microsoft Windows)
007-1680-100
111
4: Database Traversal
Small-Feature Culling Using Cull Programs
In small-feature culling, geometry that is smaller (in screen space) than a given number
of pixels is culled. Since cull programs have access to bounding spheres of nodes and
bounding boxes of geosets, the cull programs can be used for small-feature culling.
A cull program that performs small-feature culling looks like the following:
PFCULLPG_TEST_SMALLER_THAN, 1.5 // smaller than 1.5 pixels
PFCULLPG_RETURN, PFCULLPG_CULL_ON_TRUE
You must specify this program both for nodes and geosets. You can find an example of
small-feature culling in the following file:
/usr/share/Performer/src/pguide/libpf/C++/cullSmallFeature.C
(IRIX and Linux)
%PFROOT%\Src\pguide\libpf\C++\cullSmallFeature.C
(Microsoft Windows)
This feature can be also accessed in Perfly from the FOV pane.
pfNode Draw Mask
Each node in the database hierarchy can be assigned a mask that dictates whether the
node is added to the display list and thereby whether it is drawn. This mask is called the
draw mask (even though it is evaluated in the cull traversal) because it tells the cull
process whether the node is drawable or not.
The draw mask of a node is set with pfNodeTravMask(). The channel also has a draw
mask, which you set with pfChanTravMask(). By default, the masks are all 1’s or
0xffffffff.
Before testing a node for visibility, the cull traversal ANDs the two masks together. If the
result is zero, the cull prunes the node. If the result is nonzero, the cull proceeds normally.
Mask testing occurs before all visibility testing and function callbacks.
Masks allow you to draw different subgraphs of the scene on different channels, to turn
portions of the scene graph on and off, or to ignore hidden portions of the scene graph
while drawing but make them active during intersection testing.
112
007-1680-100
Controlling and Customizing Traversals
pfNode Cull and Draw Callbacks
One of the primary mechanisms for extending OpenGL Performer is through the use of
function callbacks, which can be specified on a per-node basis. OpenGL Performer
allows separate cull and draw callbacks, which are invoked both before and after node
processing. Node callbacks are set with pfNodeTravFuncs().
Cull callbacks can direct the cull traversal, while draw callbacks are added to the display
list and later executed in the draw traversal for custom rendering. There are pre-cull and
pre-draw callbacks, invoked before a node is processed, and post-cull and post-draw
callbacks, invoked after the node is processed.
The cull callbacks return a value indicating how the cull traversal should proceed, as
shown in Table 4-6.
Table 4-6
Cull Callback Return Values
Value
Action
PFTRAV_CONT
Continue and traverse the children of this node.
PFTRAV_PRUNE
Skip the subgraph rooted at this node and continue.
PFTRAV_TERM
Terminate the entire traversal.
Callbacks are processed by the cull traversal in the following order:
1.
If a pre-cull callback is defined, then call the pre-cull callback to get a cull result and
find out whether traversal should continue. Possible return values are listed in
Table 4-6.
2. If the pre-cull callback returns PFTRAV_PRUNE, the traversal returns to the parent
and continues with the node’s siblings, if any. If the callback returns
PFTRAV_TERM, the traversal terminates immediately. Otherwise, cull processing
continues.
3. If the pre-cull callback does not set the cull result using pfCullResult(), and
view-frustum culling is enabled, then perform the standard node-within-frustum
test and set the cull result accordingly.
4. If the cull result is PFIS_FALSE, skip the traversal of children. The post-cull callback
is invoked and traversal returns so that the parent node can traverse any siblings.
007-1680-100
113
4: Database Traversal
5. If a pre-draw callback is defined, then place a libpr display-list packet in the
display list so that the node’s pre-draw callback will be called by the draw process.
If running a combined CULLDRAW traversal, invoke the pre-draw callback directly
instead.
6. Process the node, continuing the cull traversal with each of the node’s children or
adding the node’s geometry to a display list (for pfGeodes). If the cull result was
PFIS_ALL_IN, view-frustum culling is disabled during the traversal of the children.
7. If a post-draw callback is defined, then place a libpr display-list packet in the
display list so that the node’s post-draw callback will be called by the draw process.
If running a combined CULLDRAW traversal, invoke the post-draw callback
directly instead.
8. If a post-cull callback is defined, then call the post-cull callback.
Draw callbacks are commonly used to place tags or change state while a subgraph is
rendered. Note that if the pre-draw callback is called, the post-draw callback is
guaranteed to be invoked. This way the callback can restore any state modified by the
pre-draw callback. This is useful for state changes such as pfPushMatrix() and
pfPopMatrix(), as shown in the environment-mapping code that is part of Example 4-2.
For doing customized culling, the pre-cull callback can determine whether a
PFIS_ALL_IN has already turned off view-frustum culling using
pfGetParentCullResult(), in which case it may not wish to do its own cull testing. It can
also find out the result of the standard cull test by calling pfGetCullResult().
Cull callbacks can also be used to render geometry (pfGeoSets) or change graphics state.
Any libpr drawing commands are captured in a display list and are later executed
during the draw traversal (see “Display Lists” in Chapter 12). However, direct graphics
library calls can be made safely only in draw function callbacks, because only the draw
process of multiprocess OpenGL Performer configurations is known to be associated
with a window.
Example 4-2 shows some sample node callbacks.
Example 4-2
pfNode Draw Callbacks
void
LoadScene(char *filename)
{
pfScene *scene = pfNewScene();
pfGroup *root = pfNewGroup();
pfGroup *reflectiveGeodes = NULL;
114
007-1680-100
Controlling and Customizing Traversals
root = pfdLoadFile(filename);
...
reflectiveGeodes =
ReturnListofGeodesWithReflectiveMaterials(root);
/* Use a node callback in the Draw process to turn on
* and off graphics library environment mapping before
* and after drawing all of the pfGeodes that have
* pfGeoStates with reflective materials.
*/
pfNodeTravFuncs(reflectiveGeodes, PFTRAV_DRAW,
pfdPreDrawReflMap, pfdPostDrawReflMap);
}
/* This callback turns on graphics library environment
* mapping. Because it changes graphics state it must be a
* Draw process node callback. */
/*ARGSUSED*/
int
pfdPreDrawReflMap(pfTraverser *trav, void *data)
{
glTexGenf(GL_S, GL_TEXTURE_GEN_MODE, GL_SPHERE_MAP);
glTexGenf(GL_T, GL_TEXTURE_GEN_MODE, GL_SPHERE_MAP);
glEnable(GL_TEXTURE_GEN_S);
glEnable(GL_TEXTURE_GEN_T);
return NULL;
}
/* This callback turns off graphics library environment
* mapping. Because it also changes graphics state it also
* must be a Draw process node callback. Also notice that
* it is important to return the graphics library’s state to
* the state at which it was in before the preNode callback
* was even made.
*/
/*ARGSUSED*/
int
pfdPostDrawReflMap(pfTraverser *trav, void *data)
{
glDisable(GL_TEXTURE_GEN_S);
glDisable(GL_TEXTURE_GEN_T);
return NULL;
}
007-1680-100
115
4: Database Traversal
Process Callbacks
The libpf library processes a visual database with a software-rendering pipeline
composed of application, cull, and draw stages. The system of process callbacks allows you
to insert your own custom culling and drawing functions into the rendering pipeline.
Furthermore, these callbacks are invoked by the proper process when your OpenGL
Performer application is configured for multiprocessing.
By default, OpenGL Performer culls and draws all active pfChannels when pfFrame() is
called. However, you can specify cull and draw function callbacks so that pfFrame() will
cause OpenGL Performer to call your custom functions instead. These functions have the
option of using the default OpenGL Performer processing in addition to their own
custom processing.
When multiprocessing is used, the rendering pipeline works on multiple frames at once.
For example, when the draw process is rendering frame n, the cull process is working on
frame n+1, and the application process is working on frame n+2. This situation requires
careful management of data so that data generated by the application is propagated to
the cull process and then to the draw process at the right time. OpenGL Performer
manages data that is passed to the process callbacks to ensure that the data is
frame-coherent and is not corrupted.
Example 4-3 illustrates the use of a cull-process callback.
Example 4-3
Cull-Process Callbacks
InitChannels()
{
...
/* create and configure all channels*/
...
/* define callbacks for cull and draw processes */
pfChanTravFunc(chan, PFTRAV_CULL, CullFunc);
pfChanTravFunc(chan, PFTRAV_DRAW, DrawFunc);
...
}
/* The Cull callback. Any work that needs to be done in the
* Cull process should happen in this function.
*/
void
CullFunc(pfChannel * chan, void *data)
{
116
007-1680-100
Process Callbacks
static long first = 1;
/* Lock down whatever processor the cull is using when
* the cull callback is first called.
*/
if (first)
{
if ((pfGetMultiprocess() & PFMP_FORK_CULL) &&
(ViewState->procLock & PFMP_FORK_CULL))
pfuLockDownCull(pfGetChanPipe(chan));
first = 0;
}
/* User-defined pre-cull processing. Application* specific cull knowledge might be used to provide
* things like line-of-sight culling.
*/
PreCull(chan, data);
/* standard Performer culling to the viewing frustum */
pfCull();
/* User-defined post-cull processing; this routine might
* be used to do things like record cull state from this
* cull to be used in future culls.
*/
PostCull(chan, data);
}
/* The draw function callback.
* Any graphics library functionality outside
* OpenGL Performer must be done here.
*/
void
DrawFunc(pfChannel *chan, void *data)
{
/* pre-Draw tasks like clearing the viewport */
PreDraw(chan, data);
pfDraw();
/* render the frame */
/* draw HUD and so on */
PostDraw(chan, data);
}
007-1680-100
117
4: Database Traversal
Process Callbacks and Passthrough Data
Cull and draw callbacks are specified on a per-pfChannel basis using the functions
pfChanTravFunc() with PFTRAV_CULL and PFTRAV_DRAW, respectively.
pfAllocChanData() allocates passthrough data, data which is passed down the rendering
pipeline to the callbacks.
In the cull phase of the rendering pipeline, OpenGL Performer invokes the cull callback
with a pointer to the pfChannel that is being culled and a pointer to the pfChannel’s
passthrough data buffer. The cull callback may modify data in the buffer. The potentially
modified buffer is then copied and passed to the user’s draw callback.
Default OpenGL Performer processing is triggered by pfCull() and pfDraw(). By default,
pfFrame() calls pfCull() first, then calls pfDraw(). If process callbacks are defined,
however, pfCull() and pfDraw() are not invoked automatically and must be called by the
callbacks to use OpenGL Performer’s default processing. pfCull() should be called only
in the cull callback; it causes OpenGL Performer to cull the current channel and to
generate a display list suitable for rendering.
Channels culled by pfCull() may be drawn in the draw callback by pfDraw(). It is valid
for the draw callback to call pfDraw() more than once. Multipass renderings performed
with multiple calls to pfDraw() are typical when you use accumulation buffer
techniques.
When the draw callback is invoked, the window will have already been properly
configured for drawing the pfChannel. Specifically, the viewport, perspective, and
viewing matrices are set to their correct values. User modifications of these values are not
reset by pfDraw(). If a draw callback is specified, OpenGL Performer does not
automatically clear the viewport; it leaves that responsibility to the application.
pfClearChan() can be called from the draw callback to clear the channel viewport. If chan
has a pfEarthSky(), then the pfEarthSky() is drawn. Otherwise, the viewport is cleared
to black and the z-buffer is cleared to its maximum value.
You should call pfPassChanData() to indicate that user data should be passed through
the rendering pipeline, which propagates the data downstream to cull and draw
callbacks. The next call to pfFrame() copies the channel buffer into internal buffers, so
that the application is then free to modify data in the buffer without fear of corruption.
The pfPassChanData() function should be called only when necessary, since calling it
imposes some buffer-copying overhead. In addition, passthrough data should be as
small as possible to reduce the time spent copying data.
118
007-1680-100
Process Callbacks
The code fragment in Example 4-4 is an example of cull and draw callbacks and the
passthrough data that is used to communicate with them.
Example 4-4
Using Passthrough Data to Communicate with Callback Routines
typedef struct
{
long val;
}
PassData;
void cullFunc(pfChannel *chan, void *data);
void drawFunc(pfChannel *chan, void *data);
int main()
{
PassData
*pd;
/* allocate passthrough data */
pd = (PassData*)pfAllocChanData(chan,sizeof(PassData));
/* initialize channel callbacks */
pfChanTravFunc(chan, PFTRAV_CULL, cullFunc);
pfChanTravFunc(chan, PFTRAV_DRAW, drawFunc);
/* main simulation loop */
while (1)
{
pfSync();
pd->val = 0;
pfPassChanData(chan);
pfFrame();
}
}
void
cullFunc(pfChannel *chan, void *data)
{
PassData
*pd = (PassData*)data;
pd->val++;
pfCull();
}
void
drawFunc(pfChannel *chan, void *data)
007-1680-100
119
4: Database Traversal
{
PassData
*pd = (PassData*)data;
fprintf(stderr, "%ld\n", pd->val);
pfClearChan(chan);
pfDraw();
}
This example would, regardless of the multiprocessing mode, have the values 0, 1, and 1
for pd->val at the points where pfFrame(), pfCull(), and pfDraw() are called. In this way,
control data can be sent down the pipeline from the application, through the cull, and on
to the draw process with frame synchronization without regard to the active
multiprocessing mode.
When configured as a process separate from the draw, the cull callback should not
attempt to send graphics commands to an OpenGL Performer window because only the
draw process is attached to the window. Callbacks should not modify the OpenGL
Performer database, but they can use pfGet() routines to inquire about database
information. The draw callback should not call glXSwapBuffers() because OpenGL
Performer must control buffer swapping in order to manage the necessary frame and
channel synchronization. However, if you need special control over buffer swapping, use
pfPipeSwapFunc() to register a function as the given pipe’s buffer-swapping function.
Once your function is registered, it will be called instead of glXSwapBuffers().
Intersection Traversal
You can make spatial inquiries in OpenGL Performer by testing the intersection of line
segments with geometry in the database. For example, a single line segment pointing
straight down from the eyepoint can determine your height above terrain, four such
segments can simulate the four tires of a car, and segments swept out by points on a
moving object can determine collisions with other objects.
Testing Line Segment Intersections
The testing of each line segment or group of spatially grouped segments requires a
traversal of part or all of a scene graph. You make these inquiries using
pfNodeIsectSegs(), which intersects the specified group of segments with the subgraph
rooted at the specified node. pfChanNodeIsectSegs() functions similarly, but includes a
channel so that the traversal can make decisions based on the level-of-detail specified by
pfLOD nodes.
120
007-1680-100
Intersection Traversal
Intersection Requests: pfSegSets
A pfSegSet is a structure that embodies an intersection request.
typedef struct _pfSegSet
{
long
mode;
void* userData;
pfSeg segs[PFIS_MAX_SEGS];
ulong activeMask;
ulong isectMask;
void* bound;
long
(*discFunc)(pfHit*);
} pfSegSet;
The segs field is an array of line segments making up the query. You tell
pfNodeIsectSegs() which segments to test with by setting the corresponding bit in the
activeMask field. If your pfSegSet contains many closely-grouped line segments, you can
specify a bounding volume using the data structure’s bound field. pfNodeIsectSegs() can
use that bounding volume to more quickly test the request against bounding volumes in
the scene graph. The userData field is a pointer with which you can point to other
information about the request that you might access in a callback. The other fields are
described in the following sections. The pfSegSet is not modified during the traversal.
Intersection Return Data: pfHit Objects
Intersection information is returned in pfHit objects. These can be queried using
pfQueryHit() and pfMQueryHit(). Table 4-7 lists the items that can be queried from a
pfHit object.
Table 4-7
007-1680-100
Intersection-Query Token Names
Query Token
Description
PFQHIT_FLAGS
Status and validity information
PFQHIT_SEGNUM
Index of the segment in a pfSegSet request
PFQHIT_SEG
Line segment as currently clipped
PFQHIT_POINT
Intersection point in object coordinates
PFQHIT_NORM
Geometric normal of an intersected triangle
121
4: Database Traversal
Table 4-7
Intersection-Query Token Names (continued)
Query Token
Description
PFQHIT_VERTS
Vertices of an intersected triangle
PFQHIT_TRI
Index of an intersected triangle
PFQHIT_PRIM
Index of an intersected primitive in pfGeoSet
PFQHIT_GSET
pfGeoSet of an intersection
PFQHIT_NODE
pfGeode of an intersection
PFQHIT_NAME
Name of pfGeode
PFQHIT_XFORM
Current transformation matrix
PFQHIT_PATH
Path in scene graph of intersection
The PFQHIT_FLAGS field is bit vector with bits that indicate whether an intersection
occurred and whether the point, normal, primitive and transformation information is
valid. For some types of intersections only some of the information has meaning; for
instance, for a pfSegSet bounding volume intersecting a pfNode bounding sphere, the
point information may not be valid.
Queries can be performed singly by calling pfQueryHit() with a single query token, or
several at a time by using pfMQueryHit() with an array of tokens. In the latter case, the
return information is placed in the specified order into a return array.
Intersection Masks
Before using pfNodeIsectSegs() to intersect the geometry in the scene graph, you must
set intersection masks for the nodes in the scene graph and correspondingly in your
search request.
Setting the Intersection Mask
The pfNodeTravMask() function sets the intersection masks in a subgraph of the scene
down through GeoSets. For example:
pfNodeTravMask(root, PFTRAV_ISECT, 0x01,
PFTRAV_SELF | PFTRAV_DESCEND, PF_SET)
122
007-1680-100
Intersection Traversal
This function sets the intersection mask of all nodes and GeoSets in the scene graph to
0x01. A subsequent intersection request would then use 0x01 as the mask in
pfNodeIsectSegs(). A description of how to use this mask follows.
Specifying Different Classes of Geometry
Databases can contain different classes of objects, and only some of those may be relevant
for a particular intersection request. For example, the wheels on a truck follow the
ground, even through a small pond; therefore, you only want to test for intersection with
the ground and not with the water. For a boat, on the other hand, intersections with both
water and the lake bottom have significance.
To accommodate distinctions between classes of objects, each node and GeoSet in a scene
graph has an intersection mask. This mask allows traversals, such as intersections, to
either consider or ignore geometry by class.
For example, you could use four classes of geometry to control tests for collision
detection of a moving ship, collision detection for a falling bowling ball, and line-of-sight
visibility. Table 4-8 matches database classes with the pfNodeTravMask() and
pfGSetIsectMask() values used to support the traversal tests listed above.
Table 4-8
Database Classes and Corresponding Node Masks
Database Class
Node Mask
Water
0x01
Ground
0x02
Pier
0x04
Clouds
0x08
Once the mask values at nodes in the database have been set, intersection traversals can
be directed by them. For example, the line segments for ship collision detection should
be sensitive to the water, ground, and pier, while those for a bowling ball would ignore
intersections with water and the clouds, testing only against the ground and pier.
Line-of-sight ranging should be sensitive to all the geometry in the scene. Table 4-9 lists
007-1680-100
123
4: Database Traversal
the traversal mask values and mask representations that would achieve the proper
intersection tests.
Table 4-9
Representing Traversal Mask Values
Intersection Class
Mask Value
Mask Representation
Ship
0x07
(Water | Ground | Pier)
Bowling ball
0x06
(Ground | Pier)
Line-of-sight ranging
0x0f
(Water | Ground | Pier | Clouds)
The intersection traversal prunes a node as soon as it gets a zero result from doing a
bitwise AND of the node intersection mask and the traversal mask specified by the
pfSegSet’s isectMask field. Thus, all nodes in the scene graph should normally be set to be
the bitwise OR of the masks of their children. After setting the class-specific masks for
different subgraphs of the scene, this can be accomplished by calling this function:
pfNodeTravMask(root, PFSET_OR, PFTRAV_SET_FROM_CHILD, 0x0);
This function sets each node’s mask by ORing 0x0 with the current mask and the masks
of the node’s children.
Note that this traversal, like that used to update node bounding volumes, is unusual in
that it propagates information up the graph from leaf nodes to root.
Discriminator Callbacks
If you need to make a more sophisticated discrimination than node masks allow about
when an intersection is valid, OpenGL Performer can issue a callback on each successful
intersection and let you decide whether the intersection is valid in the current context.
If a callback is specified in pfNodeIsectSegs(), then at each level where an intersection
occurs—for example, with bounding volumes of libpf pfGeodes (mode
PFTRAV_IS_GEODE), libpr GeoSets (mode PFTRAV_IS_GSET), or individual
geometric primitives (mode PFTRAV_IS_PRIM)—OpenGL Performer invokes the
callback, giving it information about the candidate intersection. The value you return
from the callback determines whether the intersection should be ignored and how the
intersection traversal should proceed.
124
007-1680-100
Intersection Traversal
If the return value includes the bit PFTRAV_IS_IGNORE, the intersection is ignored. The
intersection traversal itself can also be influenced by the callback. The traversal is subject
to three possible fates, as detailed in Table 4-10.
Table 4-10
Possible Traversal Results
Set Bits
Meaning
PFTRAV_CONT
Continue the traversal inside this subgraph or GeoSet.
PFTRAV_PRUNE
Continue the traversal but skip the rest of this subgraph or GeoSet.
PFTRAV_TERM
Terminate the traversal here.
Line Segment Clipping
Usually, the intersection point of most interest is the one that is nearest to the beginning
of the segment. By default, after each successful intersection, the end of the segment is
clipped so that the segment now ends at the intersection point. Upon the final return
from the traversal, it contains the closest intersection point.
However, if you want to examine all intersections along a segment you can use a
discriminator callback to tell OpenGL Performer not to clip segments—simply leave out
the PFTRAV_IS_CLIP_END bit in the return value. If you want the farthest intersection
point, you can use PFTRAV_IS_CLIP_START so that after each intersection the new
segment starts at the intersection point and extends outward.
Traversing Special Nodes
Level-of-detail nodes are intersected against the model for range zero, which is typically
the highest level-of-detail (LOD). If you want to select a different model, you can turn off
the intersection mask for the LOD node and place a switch node in parallel (having the
same parent and children as the LOD) and set it to the desired model.
Sequences and switches intersect using the currently active child or children. Billboards
are not intersected, since no eyepoint is defined for intersection traversals.
007-1680-100
125
4: Database Traversal
Picking
The pfChanPick() function provides a simple interface for intersection testing by
enabling the user to move a mouse to select one or more geometries. The method uses
pfNodeIsectSegs() and uses the high bit, PFIS_PICK_MASK, of the intersection mask in
the scene graph. Setting up picking with pfNodePickSetup() sets this bit in the
intersection mask throughout the specified subgraph but does not enable caching inside
pfGeoSets. See “Performance” on page 126.
The pfChanPick() function has an extra feature: it can either return the closest
intersection (PFPK_M_NEAREST) or return all pfHits along the picking ray
(PFPK_M_ALL).
Performance
The intersection traversal uses the hierarchical bounding volumes in the scene graph to
allow culling of the database and then processes candidate GeoSets by testing against
their internal geometry. For this reason, the hierarchy should reflect the spatial
organization of the database. High-performance culling has similar requirements (see
Chapter 24, “Performance Tuning and Debugging”).
Performance Trade-offs
OpenGL Performer currently retains no information about spatial organization of data
within GeoSets; so, each triangle in the GeoSet must be tested. Although large GeoSets
are good for rendering performance in the absence of culling, spatially localized GeoSets
are best for culling (since a GeoSet is the smallest culling unit), and spatially localized
GeoSets with few primitives are best for intersections.
Front Face/Back Face
One way to speed up intersection testing is to turn on PFTRAV_IS_CULL_BACK. When
this flag is enabled, only front-facing geometry is tested.
Enabling Caching
Precomputing information about normals and projections speeds up intersections inside
GeoSets. For the best performance, you should enable caching in GeoSets when you set
the intersection masks with pfNodeTravMask().
126
007-1680-100
Intersection Traversal
If the geometry within a GeoSet is dynamic, such as waves on a lake, caching can cause
incorrect results. However, for geometry that changes only rarely, you can use
pfGSetIsectMask() to recompute the cache as needed.
Intersection Methods for Segments
Normally, when intersecting down to the primitive level each line segment is separately
tested against each bounding volume in the scene graph, and after passing those tests is
intersected against the pfGeoSet bounding box. Segments that intersect the bounding
box are eventually tested against actual geometry.
When a pfSegSet has a spatially localized group of at least several line segments, you can
speed up the traversal by providing a bounding volume. You can use
pfCylAroundSegs() to create a bounding cylinder for the segments, place a pointer to the
resulting cylinder in the pfSegSet’s bound field, then OR the PFTRAV_IS_BCYL bit into
the pfSegSet’s mode field.
If only a rough volume-volume intersection is required, you can specify a bounding
cylinder in the pfSegSet without any line segments at all and request discriminator
callbacks at the PFTRAV_IS_NODE or PFTRAV_IS_GSET level.
Figure 4-4 illustrates some aspects of this process. The portion of the figure labeled A
represents a single segment; B is a collection of nonparallel segments, not suitable for
tightly bounding with a cylinder; and C shows parallel segments surrounded by a
bounding cylinder. In the bottom portion of the figure, the bounding cylinder around the
segments intersects the bounding box around the object; each segment in the cylinder,
thus, must be tested individually to see if any of them intersect.
007-1680-100
127
4: Database Traversal
Figure 4-4
128
Intersection Methods
007-1680-100
Chapter 5
5. Frame and Load Control
This chapter describes how to manage the display operations of a visual simulation
application to maintain the desired frame rate and visual performance level. In addition
this chapter covers advanced topics including multiprocessing and shared memory
management.
Frame Rate Management
A frame is the period of time in which all processing must be completed before updating
the display with a new image, for example, a frame rate of 60 Hz means the display is
updated 60 times per second and the time extent of a frame is 16.7 milliseconds. The
ability to fit all processing within a frame depends on several variables, some of which
are the following:
•
The number of pixels being filled
•
The number of transformations and modal changes being made
•
The amount of processing required to create a display list for a single frame
•
The quantity of information being sent to the graphics subsystem
Through intelligent management of SGI CPU and graphics hardware, OpenGL
Performer minimizes the above variables in order to achieve the desired frame rate.
However, in some cases, peak frame rate is less important than a fixed frame rate. Fixed
frame rate means that the display is updated at a consistent, unvarying rate. While a
simple step toward achieving a fixed frame rate is to reduce the maximum frame rate to
an easily achievable level, we shall explore other (less Draconian) mechanisms in this
chapter that do not adversely impact frame rates.
As discussed in the following sections, OpenGL Performer lets you select the frame rate
and has built-in functionality to maintain that frame rate and control overload situations
when the draw time exceeds or grows uncomfortably close to a frame time. While these
methods can be effective, they do require some cooperation from the run-time database.
007-1680-100
129
5: Frame and Load Control
In particular, databases should be modeled with levels-of-detail and be spatially
arranged.
Selecting the Frame Rate
OpenGL Performer is designed to run at the fixed frame rate as specified by
pfFrameRate(). Selecting a fixed frame rate does not in itself guarantee that each frame
can be completed within the desired time. It is possible that some frames might require
more computation time than is allotted by the frame rate. By taking too long, these
frames cause dropped or skipped frames. A situation in which frames are dropped is called
an overload or overrun situation. A system that is close to dropping frames is said to be in
stress.
Achieving the Frame Rate
The first step towards achieving a frame rate is to make sure that the scene can be
processed in less than a frame’s time—hopefully much less than a frame’s time.
Although minimizing the processing time of a frame is a huge effort, rife with tricks and
black magic, certain techniques stand out as OpenGL Performer’s main weapons against
slothful performance:
•
Multiprocessing. The use of multiple processes on multi-CPU systems can
drastically increase throughput.
•
View culling. By trivially rejecting portions of the database outside the viewing
volume, performance can be increased by orders of magnitude.
•
State sorting. Many graphics pipelines are sensitive to graphics mode changes.
Sorting a scene by graphics state greatly reduces mode changes, increasing the
efficiency of the hardware.
•
Level-of-detail. Objects that are far away project to a relatively small area of the
display so fewer polygons can be used to render the object without substantial loss
of image quality. The overall result is fewer polygons to draw and improved
performance.
Multiprocessing and level-of-detail is discussed in this chapter while view culling and
state sorting are discussed in Chapter 4, “Database Traversal.” More information on
sorting in the context of performance tuning can be found in Chapter 24, “Performance
Tuning and Debugging.”
130
007-1680-100
Frame Rate Management
Fixing the Frame Rate
Frame intervals are fixed periods of time but frame processing is variable in nature.
Because things change in a scene, such as when objects come into the field of view, frame
processing cannot be fixed. In order to maintain a fixed frame rate, the average frame
processing time must be less than the frame time so that fluctuations do not exceed the
selected frame rate. Alternately, the scene complexity can be automatically reduced or
increased so that the frame rate stays within a user-defined “sweet spot.” This
mechanism requires that the scene be modeled with levels of detail (pfLOD nodes).
OpenGL Performer calculates the system load for each frame. Load is calculated as the
percentage of the frame period it took to process the frame. Then if the default
OpenGL Performer fixed frame rate mechanisms are enabled, load is used to calculate
system stress, which is in turn used to adjust the level of detail (LOD) of visible models.
LOD management is OpenGL Performer’s primary method of managing system load.
Table 5-1 shows the OpenGL Performer functions for controlling frame processing.
Table 5-1
Frame Control Functions
Function
Description
pfFrameRate()
Set the desired frame rate.
pfSync()
Synchronize processing to frame boundaries.
pfFrame()
Initiate frame processing.
pfPhase()
Control frame boundaries.
pfChanStressFilter()
Control how stress is applied to LOD ranges.
pfChanStress()
Manually control the stress value.
pfGetChanLoad()
Determine the current system load.
pfChanLODAttr()
Control how LOD is performed, including global LOD adjustment and
blending (fade).
Figure 5-1 shows a frame-timing diagram that illustrates what occurs when frame
computations are not completed within the required interval. The solid vertical lines in
Figure 5-1 represent frame-display intervals. The dashed vertical lines represent video
refresh intervals.
007-1680-100
131
5: Frame and Load Control
Refresh count
modulo three
0
1
2
0
1
Overrun
2
0
1
2
0
Floating
Locked
Video
refresh
interval
Frame display interval
Time in seconds
1/60TH
1/20TH
Figure 5-1
Frame Rate and Phase Control
In this example, the video scan rate is 60 Hz and the frame rate is 20 Hz. With the video
hardware running at 60 Hz, each of the 20 Hz frames should be scanned to the video
display three times, and the system should wait for every third vertical retrace signal
before displaying the next image. The numbers across the top of the figure represent the
refresh count modulo three. New images are displayed on refreshes whose count modulo
three is zero, as shown by the solid lines.
In the first frame of this example, the new image is not yet completed when the third
vertical retrace signal occurs; therefore, the same image must be displayed again during
the next interval. This situation is known as frame overrun, because the frame
computation time extends past a refresh boundary.
Frame Synchronization
Because of the overrun, the frame and refresh interval timing is no longer synchronized;
it is out of phase. A decision must be made either to display the same image for the
remaining two intervals, or to switch to the next image even though the refresh is not
aligned on a frame boundary. The frame-rate control mode, discussed in the next section,
determines which choice is selected.
Knowing that the situation illustrated in Figure 5-1 is a possibility, you can specify a
frame control mode to indicate what you would like the system to do when a frame
overrun occurs.
132
007-1680-100
Frame Rate Management
To specify a method of frame-rate control, call pfPhase(). There are the following choices:
•
Free run without phase control (PFPHASE_FREE_RUN) tells the application to run
as fast as possible—to display each new frame as soon as it is ready, without
attempting to maintain a constant frame rate.
•
Free run without phase control but with a limit on the maximum frame rate
(PFPHASE_LIMIT) tells the application to run no faster than the rate specified by
pfFrameRate().
•
Fixed frame rate with floating phase (PFPHASE_FLOAT) allows the drawing
process to display a new frame (using glXSwapBuffers() at any time, regardless of
frame boundaries).
•
Fixed frame rate with locked phase (PFPHASE_LOCK) requires the draw process to
wait for a frame boundary before displaying a new frame.
•
The draw by default will wait for a new cull result to execute its stage functions.
This behavior can be changed by including the token PFPHASE_SPIN_DRAW with
the desired mode token from the above choices. This will allow the draw to run
every frame, redrawing the previous cull result. This can allow you to make
changes of your own in draw callback functions. Objects such as viewing frustum,
pfLODs, pfDCSs, and anything else normally processed by the cull or application
processes will not be updated until the next full cull result is available.
Free-Running Frame-Rate Control
The simplest form of frame-rate control, called free-running, is to have no control at all.
This uncontrolled mode draws frames as quickly as the hardware is able to process them.
In free-running mode, the frame rate may be 60 Hz in the areas of low database
complexity, but could drop to a slower rate in views that place greater demand on the
system. Use pfPhase(PFPHASE_FREE_RUN) to specify a free-running frame rate.
In applications in which real-time graphics provide the majority of visual cues to an
observer, the variable frame rates produced by the free-running mode may be
undesirable. The variable lag in image update associated with variable frame rate can
lead to motion sickness for the simulation participants, especially in motion
platform-based trainers or ingressive head-mounted displays. For these and other
reasons it is usually preferable to maintain a steady, consistent frame-update rate.
007-1680-100
133
5: Frame and Load Control
Fixed Frame-Rate Control
Assume that the overrun frame in Figure 5-1 completes processing during the next
refresh period, as shown. After the overrun frame, the simulation is still running at the
chosen 20-Hz rate and is updating at every third vertical retrace. If a new image is
displayed at the next refresh, its start time lags by 1/60th of a second, and therefore it is
out of phase by that much.
Subsequent images are displayed when the refresh count modulo three is one. As the
simulation continues and additional extended frames occur, the phase continues to drift.
This mode of operation is called floating phase, as shown by the frame in Figure 5-1
labeled "Floating." Use pfPhase(PFPHASE_FLOAT) to select floating-phase frame
control.
The alternative to displaying a new image out of phase is to display the old image for the
remainder of the current update period, then change to the new image at the normal
time. This locked phase extends each frame overrun to an integral multiple of the selected
frame time, making the overrun more evident but also maintaining phase throughout the
simulation. This timing is shown by the frame in Figure 5-1 labeled Locked. Although
this mode is the most restrictive, it is also the most desirable in many cases. Use
pfPhase(PFPHASE_LOCK) to select phase-locked frame control.
For example, a 20-Hz phase-locked frame rate is selected by specifying the following:
pfPhase(PFPHASE_LOCK);
pfFrameRate(20.0f);
These specifications prevent the system from switching to a newly computed image until
a display period of 1/20th second has passed from the time the previous image was
displayed. The frame rate remains fixed even when the Geometry Pipeline finishes its
work in less time. Fixed frame-rate display, therefore, involves setting the desired frame
rate and selecting one of the two fixed-frame-rate control modes.
Frame Skipping
When multiple frame times elapse during the rendering of a single frame, the system
must choose which frame to draw next. If the per-frame display lists are processed in
strict succession even after a frame overrun, the visual image slowly recedes in time and
the positional correlation between display and simulation is lost. To avoid this problem,
only the most recent frame definition received by the draw process is sent to the
134
007-1680-100
Frame Rate Management
Geometry Pipeline, and all intervening frame definitions are abandoned. This is known
as dropping or skipping frames and is performed in both of the fixed frame-rate modes.
Because the effects of variable frame rates, phase variance, and frame dropping are
distracting, you should choose a frame rate with care. Steady frame rates are achieved
when the frame time allows the worst-case view to be computed without overload. The
structure of the visual database, particularly in terms of uniform “complexity density,”
can be important in maximizing the system frame rate. See “Organizing a Database for
Efficient Culling” in Chapter 4 and Figure 4-3 for examples of the importance of database
structure.
Maintaining a fixed frame rate involves managing future system load by adjusting
graphics display actions to compensate for varying past and present loads. The theory
behind load management and suggested methods for dealing with variable load
situations are discussed in the “Level-of-Detail Management” on page 136 of this
chapter.
Sample Code
Example 5-1 demonstrates a common approach to frame control. The code is based on
part of the main.c source file used in the perfly sample application.
Example 5-1
Frame Control Excerpt
/* Set the desired frame rate. */
pfFrameRate(ViewState->frameRate);
/* Set the MP synchronization phase. */
pfPhase(ViewState->phase);
/* Application main loop */
while (!SimDone())
{
/* Sleep until next frame */
pfSync();
/* Should do all latency-critical processing between
* pfSync() and pfFrame(). Such processing usually
* involves changing the viewing position.
*/
PreFrame();
/* Trigger cull and draw processing for this frame. */
007-1680-100
135
5: Frame and Load Control
pfFrame();
/* Perform non-latency-critical simulation updates. */
PostFrame();
}
Level-of-Detail Management
All graphics systems have finite capabilities that affect the number of geometric
primitives that can be displayed per frame at a specified frame rate. Because of these
limitations, maximizing visual cues while minimizing the polygon count in a database is
often an important aspect of database development. Level-of-detail (LOD) processing is
one of the most beneficial tools available for managing database complexity for the
purpose of improving display performance.
The basic premise of LOD processing is that objects that are barely visible, either because
they are located a great distance from the eyepoint or because atmospheric conditions
reduce visibility, do not need to be rendered in great detail in order to be recognizable.
This is in stark contrast to mandating that all polygons be rendered regardless of their
contribution to the visual scene. Both atmospheric effects and the visual effect of
perspective decrease the importance of details as range from the eyepoint increases. The
predominant visual effect of distance is the perspective foreshortening of objects, which
makes them appear to shrink in size as they recede into the distance.
To save rendering time, objects that are visually less important in a frame can be rendered
with less detail. The LOD approach to optimizing the display of complex objects is to
construct a number of progressively simpler versions of an object and to select one of
them for display as a function of range.
This requires you to create multiple models of an object with varying levels of detail. You
also must supply a rule to determine how much detail is appropriate for a given distance
to the eyepoint. The sections that follow describe how to create multiple LOD models
and how to control when the changeover to a different LOD occurs.
Level-of-Detail Models
Most objects comprise smaller objects that become visually insignificant at ranges where
the conglomerate object itself is still quite prominent. For example, a complex model of
136
007-1680-100
Level-of-Detail Management
an automobile might have door handles, side- and rear-view mirrors, license plates, and
other small details.
A short distance away, these features may no longer be visible, even though the car itself
is still a visually significant element of the scene. It is important to realize that as a group,
these small features may contain as many polygons as the larger car itself, and thus have
a detrimental effect on rendering speed.
You can construct two LOD models simply by providing one model that contains all of
the detailed features and another model that contains only the car body itself and none
of the detailed features. A more sophisticated scheme uses multiple LOD models that are
grouped under an LOD node.
Figure 5-2 shows an LOD node with multiple children numbered 1 through n. In this
case, the model named LOD 1 is the most detailed model and models LOD 2 through
LOD n represent progressively coarser models. Each of these LOD models might contain
children that also have LOD components. Associated with the LOD node is a list of
ranges that define the distance at which each model is appropriate to display. There is no
limit to the number of levels of detail that can be used.
Level
of Detail
Node
LOD 1
Figure 5-2
LOD 2
LOD n
Level-of-Detail Node Structure
The object can be transformed as needed. During the culling phase of frame processing,
the distance from the eyepoint to the object is computed and used (with other factors) to
select which LOD model to display.
007-1680-100
137
5: Frame and Load Control
The OpenGL Performer pfLOD node contains a value known as the center of LOD
processing. The LOD center point is an x, y, z location that defines the point used in
conjunction with the eyepoint for LOD range-switching calculations, as described in the
section “Level-of-Detail Range Processing” on page 141 of this chapter.
Figure 5-3 shows an example in which multiple LOD models grouped under a parent
LOD node are used to represent a toy race car.
138
007-1680-100
Level-of-Detail Management
Blend
zones
LOD n
LOD 2
LOD 1
Switch
ranges
Figure 5-3
Level-of-Detail Processing
Figure 5-3 demonstrates that each car in a row of identical cars placed at increasing range
from the eyepoint is drawn using a different child of the tree’s LOD node.
007-1680-100
139
5: Frame and Load Control
The double-ended arrows indicate a switch range for each level of detail. When the car
is closer to the eyepoint than the first range, nothing is drawn. When the car is between
the first and second ranges, LOD 1 is drawn. When the car is between the second and
third ranges, LOD 2 is drawn.
This range bracketing continues until the final range is passed, at which point nothing is
drawn. The pfLOD node’s switch range list contains one more entry than the number of
child nodes to allow for this range bracketing.
OpenGL Performer provides the ability to specify a blend zone for each switch between
LOD models. These blend zones will be discussed in more detail in “Level-of-Detail
Transition Blending” on page 145.
Level-of-Detail States
In addition to standard LOD nodes, OpenGL Performer also supports LOD state—the
pfLODState. A pfLODState is in essence a way of creating classes or priorities among
LODs. A pfLODState contains eight parameters used to modify four different ways in
which OpenGL Performer calculates LOD switch ranges and LOD transition distances.
LOD states contain the following parameters:
•
Scale for LODs switch ranges
•
Offset for LODs switch ranges
•
Scale for the effect of Stress of switch ranges
•
Offset for the effect of Stress on switch ranges
•
Scale for the transition distances per LOD switch
•
Offset for the transition distances per LOD switch
•
Scale for the effect of stress on transition distances
•
Offset for the effect of stress on transition distances
These LOD states can then be attached to either single or multiple LOD nodes such that
the LOD behavior of groups or classes of objects can be different and be easily modified.
The man pages for pfLODLODState() and pfLODLODStateIndex() contain detailed
information on how to attach pfLODStates.
140
007-1680-100
Level-of-Detail Management
LOD states are useful because in a particular scene there often exists an object of focus
such as a sign, a target, or some other object of particular visual significance that needs
to be treated specially with regard to visual importance and thus LOD behavior. It stands
to reason that this particular object (or small group of objects) should be at the highest
detail possible despite being farther away than other elements in the scene which might
not be as visually significant. In fact, it might be feasible to diminish the detail of less
important objects (like rocks and trees) in favor of the other more important objects
(despite these objects being more distant). In this case one would create two LOD states.
The first would be for the important objects and could disable the effect of stress on these
nodes as well as scale the switch ranges such that the object(s) would maintain more
detail for further ranges. The second LOD state would be used to make the objects of less
importance be more responsive to system stress and possibly scale their switch ranges
such that they would show even less detail than normal. In this way, LOD states allow
biasing among different LODs to maintain desirable rendering speeds while maintaining
the visual integrity of various objects depending on their subjective importance (rather
than solely on their current visual significance).
In some multichannel applications, LOD states are used to control the action of LODs in
different viewing channels that have different visual significance criteria—for instance
one channel might be a normal channel while a second might represent an infrared
display. Rather than simple use of LOD states, it is also possible to specify a list of LOD
states to a channel and use indexes from this list for particular LODs (with
pfChanLODStateList() and pfLODLODStateIndex()). In this way, in the normal
channel a car’s geometry might be particularly important while in the infrared channel,
the hot exhaust of the same car might be much more important to observe. This type of
channel-dependent LOD can be set up by using two distinct and different LOD states for
the same index in the lists of LOD states specified for unique channels.
Note that because OpenGL Performer performs LOD calculations in a range squared
space as much as possible for efficiency reasons, LOD computation becomes more costly
when LOD states contain scales that are not equal to 1.0 or offsets not equal to 0.0 for
transitions or switch ranges—these offsets force OpenGL Performer to perform
otherwise avoidable square root calculations in order to correctly calculate the effects of
scale and offset on the LOD.
Level-of-Detail Range Processing
The LOD switch ranges present in LOD nodes are processed before being used to make
the level of detail selection. The goal of range setting is to switch LODs as objects reach
certain levels of perceptibility. The size of a channel in pixels, the field of view used in
007-1680-100
141
5: Frame and Load Control
viewing, and the distance from the observer to the display surface all affect object
perceptibility.
OpenGL Performer uses a channel size of 1024x1024 pixels and a 45-degree field of view
as the basis for calculating LOD switching ranges. The screen space size of a channel and
the current field of view are used to compute an LOD scale factor that is updated
whenever the channel size or the field of view changes.
There is an additional global LOD scale factor that can be used to adjust switch ranges
based on the relationship between the observer and the display surface. The default
global scale factor is 1.
Note that LOD switch ranges are also affected by LOD states that have been attached to
either a particular LOD or to a channel that contains the LOD. These LOD states provide
the mechanism to apply both a scale and an offset for an LODs switch ranges and to the
effect of system stress on those switch ranges. See “Level-of-Detail States” on page 140
for more information on pfLODStates.
Ultimately, an LOD’s switch range without regard to system stress can be computed as
follows:
switch_range[i] =
(range[i] *
LODStateRangeScale *
ChannelLODStateRangeScale +
LODStateRangeOffset +
ChannelLODStateRangeOffset) *
ChannelLODScale *
ChannelSizeAndFOVFactor;
If OpenGL Performer channel stress processing is active, the computed range is modified
as follows:
switch_range[i] *=
(ChannelLODStress *
LODStateRangeStressScale *
ChannelLODStateRangeStressScale +
LODStateRangeStressOffset +
ChannelLODStateRangeStressOffset);
Example 5-2 illustrates how to set LOD ranges.
142
007-1680-100
Level-of-Detail Management
Example 5-2
Setting LOD Ranges
/* setLODRanges() -- sets the ranges for the LOD node. The
* ranges from 0 to NumLODs are equally spaced between min
* and max. The last range, which determines how far you
* can get from the object and still see it, is set to
* visMax.
*/
void
setLODRanges(pfLOD *lod, float min, float max, float visMax)
{
int i;
float range, rangeInc;
rangeInc = (max - min)/(ViewState->shellLOD + 1);
for (range = min, i = 0; i < ViewState->shellLOD; i++)
{
ViewState->range[i] = range;
pfLODRange(lod, i, range);
range += rangeInc;
}
ViewState->range[i] = visMax;
pfLODRange(lod, i, visMax);
}
/* generateShellLODs() -- creates shell LOD nodes according
* to the parameters specified in the shared data structure.
*/
void
generateShellLODs(void)
{
int i;
pfGroup *grp;
pfVec4 clr;
long numLOD = ViewState->shellLOD;
long numPnts = ViewState->shellPnts;
long numPcs = ViewState->shellPcs;
for (i = 1; i <= numLOD; i++)
{
if (ViewState->shellColor == SHELL_COLOR_SING)
pfSetVec4(clr, 0.9f, 0.1f, 0.1f, 1.0f);
else
/* set the color. highest level = RED;
* middle LOD = GREEN; lowest LOD = BLUE
007-1680-100
143
5: Frame and Load Control
*/
pfSetVec4(clr,
(i <= (long)floor((double)(numLOD/2.0f)))?
(-2.0f/numLOD) * i + 1.0f + 2.0f/numLOD:
0.0f,
(i <= (long)floor((double)(numLOD/2)))?
(2.0f/numLOD) * (i - 1):
(-2.0f/numLOD) * i + 2.0f,
(i <= (long)floor((double)(numLOD/2)))?
0.0f:
(2.0f/numLOD) * i - 1.0f,
1.0f);
/* build a shell GeoSet */
grp = createShell(numPcs, numPnts,
ViewState->shellSweep, &clr,
ViewState->shellDraw);
normalizeNode((pfNode *)grp);
/* add geode as another level of detail node */
pfAddChild(ViewState->LOD, grp);
/* simplify the geometry, but don’t have less than
* 4 points per circle or less than 3 pieces */
numPnts = (numPnts > 7) ? numPnts-4 : 4;
numPcs = (numPcs > 6) ? numPcs-4 : 3;
}
}
...
ViewState->LOD = pfNewLOD();
generateShellLODs();
/* get the LOD’s extents */
pfGetNodeBSphere(ViewState->LOD, &(ViewState->bSphere));
pfLODCenter(ViewState->LOD, ViewState->bSphere.center);
/* set ranges for LODs; there should be (num LODs + 1)
* range entries */
setLODRanges(ViewState->LOD, ViewState->minRange,
ViewState->maxRange, ViewState->max);
144
007-1680-100
Level-of-Detail Management
Level-of-Detail Transition Blending
An undesirable effect called popping occurs when the sudden transition from one LOD to
the next LOD is visually noticeable. This distracting image artifact can be ameliorated
with a slight modification to the normal LOD-switching process.
In this modified method, a transition per LOD switch is established rather than making
a sudden substitution of models at the indicated switch range. These transitions specify
distances over which to blend between the previous and next LOD. These zones are
considered to be centered at the specified LOD switch distance, as shown by the
horizontal shaded bars of Figure 5-3. Note that OpenGL Performer limits the transition
distances to be equal to the shortest distance between the switch range and the two
neighboring switch ranges. For more information, see the pfLODTransition() man page.
As the range from eyepoint to LOD center-point transitions the blend zone, each of the
neighboring LOD levels is drawn by using transparency-to-composite samples taken
from the present LOD model with samples taken from the next LOD model. For example,
at the near, center, and far points of the transition blend zone between LOD 1 and LOD
2, samples from both LOD 1 and LOD 2 are composited until the end of the transition
zone is reached, where all the samples are obtained from LOD 2.
Table 5-2 lists the transparency factors used for transitioning from one LOD range to
another LOD range.
Table 5-2
LOD Transition Zones
Distance
LOD 1
LOD 2
Near edge of blend zone
100% opaque
0% opaque
Center of blend zone
50% opaque
50% opaque
Far edge of blend zone
0% opaque
100% opaque
LOD transitions are made smoother and much less noticeable by applying a blending
technique rather than making a sudden transition. Blending allows LOD transitions to
look good at ranges closer to the eye than LOD popping allows. Decreasing switch
ranges in this way improves the ability of LOD processing to maximize the visual impact
of each polygon in the scene without creating distracting visual artifacts.
The benefits of smooth LOD transition have an associated cost. The expense lies in the
fact that when an object is within a blend zone, two versions of that object are drawn. This
007-1680-100
145
5: Frame and Load Control
causes blended LOD transitions to increase the scene polygon complexity during the
time of transition. For this reason, the blend zone is best kept to the shortest distance that
avoids distracting LOD-popping artifacts. Currently, fade level of detail is supported
only on RealityEngine and InfiniteReality graphics systems.
Note that the actual ‘blend’ or ‘fade’ distance used by OpenGL Performer can also be
adjusted by the LOD priority structures called pfLODStates. pfLODStates hold an offset
and scale for the size of transition zones as well as an offset and scale for how system
stress can affect the size of the transition zones. See “Level-of-Detail States” on page 140
for more information on pfLODStates.
Note also, that there exists a global LOD transition scale on a per channel basis that can
affect all transition distances uniformly.
Thus for an LOD with 5 switch ranges R0, R1, R2, R3, R4 to switch between four models
(M0, M1, M2, M3), there are 5 transition zones T0 (fade in M0), T1 (blend between M0
and M1), T2 (blend between M1 and M2), T3 (blend between M2 and M3), T4 (fade out
M3). The actual fade distances (without regard to channel stress) are as follows:
fadeDistance[i] =
(transition[i] *
LODStateTransitionScale *
ChannelLODStateTransitionScale +
LODStateTransitionOffset +
ChannelLODStateTransitionOffset) *
ChannelLODFadeScale;
If OpenGL Performer management of channel stress is turned on then the above fade
distance is modified as follows:
fadeDistance[i] /=
(ChannelStress *
LODStateTransitionStressScale *
ChannelLODStateTransitionStressScale +
LODStateTransitionStressOffset +
ChannelLODStateTransitionStressOffset);
Run-Time User Control Over LOD Evaluation
A pfLOD node provides one last resort for applications that have complex level-of-detail
calculations. For example, an application might wish to limit the speed at which different
LODs of an object switch. When switching depends on the range from the camera, a very
146
007-1680-100
Level-of-Detail Management
fast-moving camera may result in rapid changes of LODs. The application may require
an artificial filter to take the simple range-based evaluation and ease it into the display
over time.
An application may take over the LOD evaluation function using the API
pfLODUserEvalFunc() on pfLOD. The user-supplied function must return a floating
point number. Similar to the result of pfEvaluateLOD(), this number picks either a single
child or a blend of two children of the pfLOD node.
Note that the performance of the cull process may decrease if the user function is too slow
to execute.
Terrain Level-of-Detail
In creating LOD models and transitions for objects, it is often safe to assume that the
entire model should transition at the same time. It is quite reasonable to make features
of an automobile such as door handles disappear from the scene at the same time even
when the passenger door is slightly closer than the driver’s door. It is much less clear that
this approach would work for very large objects such as an aircraft carrier or a space
station, and it is clearly not acceptable for objects that span a large extent, such as a terrain
surface.
Active Surface Definiton (ASD)
Attempts to handle large-extent objects with discrete LOD tools focus on breaking the big
object into myriad small objects and treating each small object independently. This works
in some cases but often fails at the junction between two or more independent objects
where cracks or seams exist when different detail levels apply to the objects. Some terrain
processing systems have attempted to provide a hierarchy of crack-filling geometry that
is enabled based on the LOD selections of two neighboring terrain patches. This “digital
grout” becomes untenable when more than a few patches share a common vertex.
You can always make the transitions between LODs smooth by using active surface
definition. ASD treats the entire terrain as a single connected surface rather than multiple
patches that are loaded into memory as necessary. The surface is modeled with several
hierarchical LOD meshes in data structures that allow for the rapid evaluation of smooth
LOD transitions, load management on the evaluation itself, and efficient generation of a
meshed terrain surface of the visible triangles for the current frame. For more
information, refer to the Chapter 20, “Active Surface Definition.”
007-1680-100
147
5: Frame and Load Control
Arbitrary Morphing
Terrain level of detail using an interpolative active surface definition is a restricted form
of the more general notion of object morphing. Morphing of models such as the car in a
previous example can simply involve scaling a small detail to a single point and then
removing it from the scene. Morphing is possible even when the topologies of
neighboring pairs do not match. Both models and terrain can have vertex, normal, color,
and appearance information interpolated between two or more representations. The
advantages of this approach include: reduced graphics complexity since blending is not
used, constant intersection truth for collision and similar tasks, and monotonic database
complexity that makes system load management much simpler. Such evaluation might
make use of the compute process and pfFlux objects to hold the vertex data and to
modify the scene graph control to chose the proper form of the object. pfSwitch nodes can
take a pfFlux for holding its value; see the pfSwitchValFlux() man page. pfLOD nodes
can take a flux for controlling range with pfLODRangeFlux(). See the pfLOD and
pfEngine man pages for more information on morphing.
Maintaining Frame Rate Using Dynamic Video Resolution
When frame rate is not maintained, some frames display longer than others. If, for
example, when the frame rate is 30 frames per second, a frame takes longer than 1/30th
of a second to fill the frame buffer, the frame is not displayed. Consequently, the current
frame is displayed for two instead of one 1/30ths of a second. The result of inconsistent
frame rates is jerky motion within the scene.
Note: You have some control over what happens when a frame rate is missed. You can
choose, for example, to begin the next frame in the next 1/60th of a second, or wait for
the start of the next 1/30th second. For more information about handling frame drawing
overruns, see pfPhase in “Free-Running Frame-Rate Control” on page 133.
The key to maintaining frame rate is limiting the amount of information to be rendered.
OpenGL Performer can take care of this problem automatically for you on InfiniteReality
systems when you use the PFPVC_DVR_AUTO token with pfPVChanDVRMode().
In PFPVC_DVR_AUTO mode, OpenGL Performer checks every rendered frame to see if
it took too long to render. If it did, OpenGL Performer reduces the size of the image, and
correspondingly, the number of pixels in it. Afterwards, the video hardware enlarges the
148
007-1680-100
Maintaining Frame Rate Using Dynamic Video Resolution
images to the same size as the pfChannel; in this way, the image is the correct size, but it
contains a reduced number of pixels, as suggested in Figure 5-4.
Figure 5-4
Real Size of Viewport Rendered Under Increasing Stress
Although the viewport is reduced as stress increases, the viewer never sees the image
grow smaller because bipolar filtering is used to enlarge the image to the size of the
channel.
The Channel in DVR
When using Dynamic Video Resolution (DVR), the origin and size of a channel are
dynamic. For example, a viewport whose lower-left corner is at the center of a pfPipe
(with coordinates 0.5, 0.5) would be changed to an origin of (0.25, 0.25) with respect to
the full pfPipe window if the DVR settings were scaled by a factors of 0.5 in both X and
Y dimensions.
If you are doing additional rendering into a pfChannel, you may need to know the size
and the actual rendered area of the pfChannel. Use pfGetChanOutputOrigin() and
pfGetChanOutputSize() to get the actual rendered origin and size, respectively, of a
pfChannel. pfGetChanOrigin() and pfGetChanSize() give the displayed origin and size
of the pfChannel and these functions should be used for mapping mouse positions or
other window-relative nonrendering positions to the pfChannel area.
Additionally, if DVR alters the rendered size of a pfChannel, a corresponding change
should be made to the width of points and lines. For example, when a channel is scaled
in size by one half, lines and points must be drawn half as wide as well so that when the
final image is enlarged, in this case by a factor of two, the lines and points scale correctly.
pfChanPixScale() sets the pixel scale factor. pfGetChanPixScale() returns this value for
a channel. pfChannels set this pixel scale automatically.
007-1680-100
149
5: Frame and Load Control
DVR Scaling
DVR scales linearly in response to the most common cause of draw overload: filling the
polygons. For example, if the DRAW stage process overran by 50%, to get back in under
the frame time, the new scene must draw 30% fewer pixels. We can do this with DVR by
rendering to a smaller viewport and letting the video hardware rescale the image to the
correct display size.
If pfPVChanMode() is set to PFPVC_DVR_AUTO, OpenGL Performer automatically
scales each of the pfChannels. pfChannels automatically scale themselves according to
the scale set on the pfPipeVideoChannel they are using.
If the pfPVChanMode() is PFPVC_DVR_MANUAL, you control scaling according to
your own policy by setting the scale and size of the pfPipeVideoChannel in the
application process between pfSync() and pfFrame(), as shown in this example:
Total pixels drawn last frame = ChanOutX * ChanOutY * Depth Complexity
To make the total pixels drawn 30% less, do the following:
NewChanOutX = NewChanOutY = .7 * (Chan OutX * ChanOut.)
New ChanOut X = sqrt (.7) * ChanOutX
New ChanOut X = sqrt (.7) * ChanOut X
NewChanOut = sqrt (.7) * ChanOut
Customizing DVR
Your application has full control over DVR behavior. You can either configure the
automatic mode or implement your own response control.
Automatic resizing can cause problems when an image has so much information in it the
viewport is reduced too drastically, perhaps to only a few hundred pixels, so that when
the image is enlarged, the image resolution is unacceptably blurry. To remedy this
problem, pfPipeVideoChannel includes the following methods to limit the reduction of
a video channel:
pfPVChanMaxDecScale()
Sets the maximum X and Y decrement scaling that can happen in a single
step of automatic dynamic video resizing. A scale value of (-1), the
default, removes the upper bound on decremental scales.
150
007-1680-100
Maintaining Frame Rate Using Dynamic Video Resolution
pfPVChanMaxIncScale()
Sets the maximum X and Y increment scaling that can happen in a single
step of automatic dynamic video resizing. A scale value of (-1), the
default, removes the upper bound on incremental scales.
pfPVChanMinDecScale()
Sets the minimum X and Y decrement scaling that can happen in a single
step of automatic dynamic video resizing. The default value is 0.0.
pfPVChanMinIncScale()
Sets the minimum X and Y increment scaling that can happen in a single
step of automatic dynamic video resizing. The default value is 0.0.
pfPVChanStress()
Sets the stress of the pfPipeVideoChannel for the current frame. This call
should be made in the application process after pfSync() and before
pfFrame() to affect the next immediate draw process frame.
pfPVChanStressFilter()
Sets the parameters for computing stress if it is not explicitly set for the
current frame by pfPVChanStress().
Each of these methods have corresponding Get methods that return the values set by
these methods.
To resize the video channel manually, use pfPipeVideoChannel sizing methods, such as
pfPVChanOutputSize(), pfPVChanAreaScale(), and pfPVChanScale().
The pfPipeVideoChannel associated with a channel is returned by pfGetChanPVChan().
If there is more than one pfPipeVideoChannel associated with a pfPipeWindow, each one
is identified by an index number. In the case of multiple pfPipeVideoChannels, the
pfPipeVideoChannel index is set using pfChanPWinPVChanIndex() and returned by
pfGetChanPWinPVChanIndex().
Understanding the Stress Filter
The pfPVChanStressFilter() function sets the parameters for computing stress for a
pfPipeVideoChannel when the stress is not explicitly set for the current frame by
pfPVChanStress(), as shown in the following example:
void pfPipeVideoChannel::setStressFilter(float *frameFrac,
float *lowLoad, float *highLoad, float *pipeLoadScale,
float *stressScale, float *maxStress);
007-1680-100
151
5: Frame and Load Control
The frameFrac argument is the fraction of a frame that pfPipeVideoChannel is expected to
take to render the frame; for example, if the rendering time is equal to the period of the
frame rate, frameFrac is 1.
If there is only one pfPipeVideoChannel, it is best if frameFrac is 1. If there are more than
one pfPipeVideoChannels on the pfPipe, by default frameFrac is divided among the
pfPipeVideoChannels. You can set frameFrac explicitly for each pfPipeVideoChannel
such that a channel rendering visually complex scenes is allocated more time than a
channel rendering simple scenes.
The pfGetPFChanStressFilter() function returns the stress filter parameters for
pfPipeVideoChannel. If stressScale is nonzero, stress is computed for the
pfPipeVideoChannel every frame. The parameters low and high define a hysteresis band
for system load. When the load is above lowLoad and below highLoad, stress is held
constant. When the load falls outside of the lowLoad and highLoad parameters,
OpenGL Performer reduces or increases stress respectively by dynamically resizing the
output area of the pfPipeVideoChannel until the load stabilizes between lowLoad and
highLoad.
If pipeStressScale is nonzero, the load of the pfPipe of the pfPipeVideoChannel are
considered in computing the stress. The parameter maxStress is the clamping value above
which the stress value cannot go. For more information about the stress filter, see the man
page for pfPipeVideoChannel.
Dynamic Load Management
Because the effects of variable image update rates can be objectionable, many simulation
applications are designed to operate at a fixed frame rate. One approach to selecting this
fixed frame rate is to select an update rate constrained by the most complex portion of
the visual database. Although this conservative approach may be acceptable in some
cases, OpenGL Performer supports a more sophisticated approach using dynamic LOD
scaling.
Using multiple LOD models throughout a database provides the traversal system with a
parameter that can be used to control the polygonal complexity of models in the scene.
The complexity of database objects can be reduced or increased by adjusting a global
LOD range multiplier that determines which LOD level is drawn.
152
007-1680-100
Dynamic Load Management
Using this facility, a closed-loop control system can be constructed that adjusts the
LOD-switching criteria based on the system load, also called stress, in order to maintain
a selected frame rate.
Figure 5-5 illustrates a stress-processing control system.
Desired Frame Time
ess
Str ers
et
ram
Pa
ter
Fil D)
s
s
O
e
Str Set L
(
sal
ver D)
a
r
T LO
e
(Us
g
rin
de
n
Re
Frame
Buffer
Actual Frame Time
Figure 5-5
Stress Processing
In Figure 5-5, the desired and actual frame times are compared by the stress filter. Based
on the user-supplied stress parameters, the stress filter adjusts the global LOD scale
factor by increasing it when the system is overloaded and decreasing it when the system
is underloaded. In this way, the system load is monitored and adjusted before each frame
is generated.
The degree of stability for the closed-loop control system is an important issue. The ideal
situation is to have a critically damped control system—that is, one in which just the right
amount of control is supplied to maintain the frame rate without introducing
undesirable effects. The effects of overdamped and underdamped systems are visually
007-1680-100
153
5: Frame and Load Control
distracting. An underdamped system oscillates, causing the system to continuously
alternate between two different LOD models without reaching equilibrium.
Overdamped systems may fail to react within the time required to maintain the desired
frame rate. In practice, though, dynamic load management works well, and simple stress
functions can handle the slowly changing loads presented by many databases.
The default stress function is controlled with user-selectable parameters. These
parameters are set using the pfChanStressFilter() function.
The default stress function is implemented by the code fragment in Example 5-3.
Example 5-3
Default Stress Function
/* current load */
curLoad = drawTime * frameRate / frameFrac;
/* integrated over time */
if (curLoad < lowLoad)
stressLevel -= stressParam * stressLevel;
else
if (curLoad > highLoad)
stressLevel += stressParam * stressLevel;
/* limited to desired range */
if (stressLevel < 1.0)
stressLevel = 1.0;
else
if (stressLevel > maxStress)
stressLevel = maxStress;
The parameters lowLoad and highLoad define a comfort zone for the control system. The
first if-test in the code fragment demonstrates that this comfort zone acts as a dead band.
Instantaneous system load within the bounds of the dead band does not result in a
change in the system stress level. If the size of the comfort zone is too small, oscillatory
distress is the probable result. It is often necessary to keep the highLoad level below the
100% point so that blended LOD transitions do not drive the system into overload
situations.
For those applications in which the default stress function is either inappropriate or
insufficient, you can compute the system stress yourself and then set the stress load
factor. Your filter function can access the same system measures that the default stress
function uses, but it is also free to keep historical data and perform any feedback-transfer
processing that application-specific dynamic load management may require.
154
007-1680-100
Successful Multiprocessing with OpenGL Performer
The primary limitation of the default stress function is that it has a reactive rather than
predictive nature. One of the major advantages of user-written stress filters is their ability
to predict future stress levels before increased or decreased load situations reach the
pipeline. Often the simulation application knows, for example, when a large number of
moving models will soon enter the viewing frustum. If their presence is anticipated, then
stress can be artificially increased so that no sudden LOD changes are required as they
actually enter the field of view.
Successful Multiprocessing with OpenGL Performer
Note: This is an advanced topic.
This section does not apply to Microsoft Windows. OpenGL Performer 3.0 for Microsoft
Windows does not support more than a single processor.
This section describes an advanced topic that applies only to systems with more than one
CPU. If you do not have a multiple-CPU system, you may want to skip this section.
OpenGL Performer uses multiprocessing to increase throughput for both rendering and
intersection detection. Multiprocessing can also be used for tasks that run
asynchronously from the main application like database management. Although
OpenGL Performer hides much of the complexity involved, you need to know
something about how multiprocessing works in order to use multiple processors well.
Review of Rendering Stages
The OpenGL Performer application renders images using one or more pfPipes as
independent software-rendering pipelines. The flow through the rendering pipeline can
be modeled using these functional stages:
007-1680-100
Intersection
Test for intersections between segments and geometry to simulate
collision detection or line-of-sight for example.
Application
Do requisite processing for the visual simulation application, including
reading input from control devices, simulating the vehicle dynamics of
moving models, updating the visual database, and interacting with
other networked simulation stations.
155
5: Frame and Load Control
Cull
Traverse the visual database and determine which portions of it are
potentially visible, perform a level-of-detail selection for models with
multiple representations, and a build sorted, optimized display list for
the draw stage.
Draw
Issue graphics library commands to a Geometry Pipeline in order to
create an image for subsequent display.
You can partition these stages into separate parallel processes in order to distribute the
work among multiple CPUs. Depending on your system type and configuration, you can
use any of several available multiprocessing models.
Choosing a Multiprocessing Model
Use pfMultiprocess() to specify which functional stages, if any, should be forked into
separate processes. The multiprocessing mode is actually a bitmask where each bit
indicates that a particular stage should be configured as a separate process. For example,
the bit PFMP_FORK_DRAW means the draw stage should be split into its own process.
Table 5-3 lists some convenient tokens that represent common multiprocessing modes.
Table 5-3
Multiprocessing Models
Model Name
Description
PFMP_APPCULLDRAW
Combine the application, cull, and draw stages into a single
process. In this model, all of the stages execute within a single
frame period. This is the minimum-latency mode of operation.
PFMP_APP_CULLDRAW
Combine the cull and draw stages in a process that is separate from
the application process. This model provides a full frame period
for the application process, while culling and drawing share this
same interval. This mode is appropriate when the host’s
simulation tasks are extensive but graphic demands are light, as
might be the case when complex vehicle dynamics are performed
but only a simple dashboard gauge is drawn to indicate the results.
or
PFMP_FORK_APP
156
007-1680-100
Successful Multiprocessing with OpenGL Performer
Table 5-3
Multiprocessing Models (continued)
Model Name
Description
PFMP_APPCULL_DRAW
Combine the application and cull stages in a process that is
separate from the draw process. This mode is appropriate for many
simulation applications when application and culling demands are
light. It allocates a full CPU for drawing and has the application
and cull stages share a frame period. Like the
PFMP_APP_CULLDRAW mode, this mode has a single frame
period of pre-draw latency.
or
PFMP_FORK_DRAW
PFMP_APP_CULL_DRAW
or
PFMP_FORK_CULL |
PFMP_FORK_DRAW
Perform the application, cull, and draw stages as separate
processes. This is the full maximum-throughput multiprocessing
mode of OpenGL Performer operation. In this mode, each pipeline
stage is allotted a full frame period for its processing. Two frame
periods of latency exist when using this high degree of parallelism.
You can also use the pfMultiprocess() function to specify the method of communication
between the cull and draw stages, using the bitmasks PFMP_CULLoDRAW and
PFMP_CULL_DL_DRAW.
Cull-Overlap-Draw Mode
Setting PFMP_CULLoDRAW specifies that the cull and draw processes for a given frame
should overlap—that is, that they should run concurrently. For this to work, the cull and
draw stages must be separate processes (PFMP_FORK_DRAW must be true). In this
mode the two stages communicate in the classic producer-consumer model, by way of a
pfDispList that is configured as a ring (FIFO) buffer; the cull process puts commands on
the ring while the draw process simultaneously consumes these commands.
The main benefit of using PFMP_CULLoDRAW is reduced latency, since the number of
pipeline stages is reduced by one and the resulting latency is reduced by an entire frame
time. The main drawback is that the draw process must wait for the cull process to begin
filling the ring buffer.
Forcing Display List Generation
When the cull and draw stages are in separate processes, they communicate through a
pfDispList; the cull process generates the display list, and the draw process traverses and
renders it. (The display list is configured as a ring buffer when using
PFMP_CULLoDRAW mode, as described in the “Cull-Overlap-Draw Mode” section).
007-1680-100
157
5: Frame and Load Control
However, when the cull and draw stages are in the same process (as occurs with the
PFMP_APPCULLDRAW or PFMP_APP_CULLDRAW multiprocessing models) a
display list is not required and by default one will not be used. Leaving out the
pfDispList eliminates overhead. When no display list is used, the cull trigger function
pfCull() has no effect; the cull traversal takes place when the draw trigger function
pfDraw() is invoked.
In some cases you may want an intermediate pfDispList between the cull and draw
stages even though those stages are in the same process. The most common situation that
calls for such a setup is multipass rendering when you want to cull only once but render
multiple times. With PFMP_CULL_DL_DRAW enabled, pfCull() generates a pfDispList
that can be rendered multiple times by multiple calls to pfDraw().
Intersection Pipeline
The intersection pipeline is a two-stage pipeline consisting of the application and the
intersection stages. The intersection stage may be configured as a separate process by
setting the PFMP_FORK_ISECT bit in the bitmask given to pfMultiprocess(). When
configured as such, the intersection process is triggered for the current frame when the
application process calls pfFrame(). Then in the special intersection callback set with
pfIsectFunc(), you can invoke any number of intersection requests with
pfNodeIsectSegs(). To support this operation, the intersection process keeps a copy of
the scene graph pfNodes.
The intersection process is asynchronous so that if it does not finish within a frame time
it does not slow down the rendering pipeline(s).
Compute Process
The compute process is an asynchronous process provided for doing extensive
asynchronous computation. The compute stage is done as part of pfFrame() in the
application process unless it is configured to run as separate process by setting the
PFMP_FORK_COMPUTE bit in the pfMultiprocess() bitmask. The compute process is
asynchronous so that if it does not finish within a frame time, it will not slow down the
rendering pipeline. The compute process is intended to work with pfFlux objects by
placing the results of asynchronous computation in pfFluxes. pfFlux will automatically
manage the needed multibuffering and frame consistency requirements for the data. See
Chapter 19, “Dynamic Data,” for more information on pfFlux. Some OpenGL Performer
objects, such as pfASD, do their computation in the compute stage so pfCompute() must
158
007-1680-100
Successful Multiprocessing with OpenGL Performer
be called from any compute user callback if one has been specified with
pfComputeFunc().
Multiple Rendering Pipelines
By default, OpenGL Performer uses a single pfPipe, which in turn draws one or more
pfChannels into one or more pfPipeWindows. If you want to use multiple rendering
pipelines, as on two- or three-Geometry Pipeline Onyx RealityEngine2 and
InfiniteReality systems, use pfMultipipe() to specify the number of pfPipes required.
When using multiple pipelines, the PFMP_APPCULLDRAW and
PFMP_APPCULL_DRAW modes are not supported and OpenGL Performer defaults to
the PFMP_APP_CULL_DRAW multiprocessing configuration. Regardless of the number
of pfPipes, there is always a single application process that triggers the rendering of all
pipes with pfFrame().
Multithreading
For additional multiprocessing and attendant increased throughput, the CULL stage of
the rendering pipeline may be multithreaded. Multithreading means that a single pipeline
stage is split into multiple processes, or threads which concurrently work on the same
frame. Use pfMultithread() to allocate a number of threads for the cull stage of a
particular rendering pipeline.
Cull multithreading takes place on a per-pfChannel basis; that is, each thread does all the
culling work for a given pfChannel. Thus, an application with only a single channel will
not benefit from multithreading the cull. An application with multiple, equally complex
channels will benefit most by allocating a number of cull threads equal to the number of
channels. However, it is valid to allocate fewer cull threads if you do not have enough
CPUs—in this case the threads are assigned to channels on a need basis.
CULL Sidekick Processes
The OpenGL Performer CULL process traverses a scene graph and culls out any invisible
geometry. Its result is a list of visible pfGeoSets. The OpenGL Performer CULL process
does not break pfGeoSets into their visible and invisible parts. This means that a
pfGeoSet whose bounding box intersects the viewing frustum will be sent to the graphics
pipe even if only one triangle in this pfGeoSet is visible.
One way to overcome this problem is to allocate extra processes for cleaning up the
pfGeoSet lists that the CULL processes produce. These extra processes are called CULL
007-1680-100
159
5: Frame and Load Control
sidekicks. By default, a CULL sidekick process checks all the primitives in all pfGeoSets
on the CULL output. It replaces original pfGeoSets with temporary pfGeoSets and
populates the temporary pfGeoSets with the visible parts of the original pfGeoSets. By
default, CULL sidekick processes test each primitive twice for the following:
•
For frustum visibility
A primitive outside the viewing frustum will be omitted from the temporary
pfGeoSet.
•
For backface culling
A primitive facing away from the viewer will be omitted from the temporary
pfGeoSet. This test is skipped when a pfGeoSet is drawn without backface testing.
Each CULL process can have multiple CULL_SIDEKICK processes. You can use the
pfMultithread() call to specify the number of CULL_SIDEKICK processes for each CULL
process. The collection of CULL_SIDEKICK processes configured for each CULL process
traverse the pfGeoSet list that the CULL process produces in a round-robin manner. The
more CULL_SIDEKICK processes (each assigned to a separate CPU), the faster they
process the pfGeoSet list that the CULL process produces. For more information about
CULL_SIDEKICK processes in the context of CULL optimizations, see section “Cull
Sidekick Processes” on page 173.
Order of Calls
The multiprocessing model set by pfMultiprocess() is used for each of the rendering
pipelines. In programs that configures the application stage as a separate process, all
OpenGL Performer calls must be made from the process that calls pfConfig() or the
results are undefined. Both pfMultiprocess(), pfMultithread(), and pfMultipipe() must
be called after pfInit() but before pfConfig(). pfConfig() configures OpenGL Performer
according to the required number of pipelines and the desired multiprocessing and
multithreading modes, forks the appropriate number of processes, and then returns
control to the application. pfConfig() should be called only once during each OpenGL
Performer application.
Comparative Structure of Models
Figure 5-6 shows timing diagrams for each of the process models. The vertical lines are
frame boundaries. Five frames of the simulation are shown to allow the system to reach
steady-state operation. Only one of these models can be selected at a time, but they are
shown together so that you can compare their structures.
160
007-1680-100
Successful Multiprocessing with OpenGL Performer
Boxes represent the functional stages and are labeled as follows:
007-1680-100
An
Application process for the nth frame
Cn
Cull process for the nth frame
Dn
Draw process for the nth frame
161
5: Frame and Load Control
APP
A
CULL
C
DRAW
D
Host
Simulation
Process
Cull
Process
(traversal)
Draw
Process
Period=1/Frame Rate
A0
PFMPAPPCULLDRAW
P0
C0
A0
D0
C0
A1
C1
A1
D1
C1
A2
C2
A2
D2
C2
A3
C3
A3
D3
C3
A4
C4
A4
D4
C4
PFMPAPPCULL_DRAW
P1
P0
A0
D0
D1
D2
D3
A1
A2
A3
A4
PFMP_APP_CULLDRAW
C0
P1
P0
A0
PFMP_APP_CULL_DRAW P1
D0
C1
A0
Start
D3
C3
A3
A4
C0
C1
C2
C3
D0
D1
D2
A2
A3
A4
C0
P2
D2
A2
A1
PFMP_APP_CULL0DRAW P1
C2
A1
P2
P0
D1
C1
D0
Frame 0
C2
D1
Frame 1
C3
D2
Frame 2
D3
Frame 3
Frame 4
Time
Figure 5-6
162
Multiprocessing Models
007-1680-100
Successful Multiprocessing with OpenGL Performer
Notice that when a stage is split into its own process, the amount of time available for all
stages increases. For example, in the case where the application, cull, and draw stages are
three separate processes, it is possible for total system performance to be tripled over the
single process configuration.
Asynchronous Database Processing
Many databases are too large to fit into main memory. A common solution to this
problem is called database paging where the database is divided into manageable chunks
on disk and loaded into main memory when needed. Usually chunks are paged in just
before they come into view and are deleted from the scene when they are comfortably
out of viewing range.
All this paging from disk and deleting from main memory takes a lot of time and is
certainly not amenable to maintaining a fixed frame rate. The solution supported by
OpenGL Performer is asynchronous database paging in which a process, completely
separate from the main processing pipeline(s), handles all disk I/O and memory
allocations and deletions. To facilitate asynchronous database paging, OpenGL
Performer provides the pfBuffer structure and the DBASE process.
DBASE Process
The database (or DBASE) process is forked by pfConfig() if the PFMP_FORK_DBASE bit
was set in the mode given to pfMultiprocess(). The database process is triggered when
the application process calls pfFrame() and invokes the user-defined callback set with
pfDBaseFunc(). The database process is totally asynchronous. If it exceeds a frame time
it does not slow down any rendering or intersection pipelines.
The DBASE process is intended for asynchronous database management when used
with a pfBuffer.
pfBuffer
A pfBuffer is a logical buffer that isolates database changes to a single process to avoid
memory collisions on data from multiple processes. In typical use, a pfBuffer is created
with pfNewBuffer(), made current with pfSelectBuffer(), and merged with the main
OpenGL Performer buffer with pfMergeBuffer(). While the DBASE process is intended
for pfBuffer use, other processes forked by the application may also use different
pfBuffers in parallel for multithreaded database management. By ensuring that only a
007-1680-100
163
5: Frame and Load Control
single process uses a given pfBuffer at a given time and following a few scoping rules
discussed in the following paragraphs, the application can safely and efficiently
implement asynchronous database paging
A pfNode is said to have buffer scope or be “in” a particular pfBuffer. This is an important
concept because it affects what you can do with a given node. A newly created node is
automatically “in” the currently active pfBuffer until that pfBuffer is merged using
pfMergeBuffer(). At that instant, the pfNode is moved into the main OpenGL Performer
buffer, otherwise known as the application buffer.
A rule in pfBuffer management is that a process may only access nodes that are in its
current pfBuffer. As a result, a database process may not directly add a newly created
subgraph of nodes to the main scene graph because all nodes in the main scene graph
have application buffer scope only—they are isolated from the database pfBuffer. This
may seem inconvenient at first but it eliminates catastrophic errors. For example, the
application process traverses a group at the same time you add a child; this changes its
child list and causes the traversal to chase a bad pointer.
Remedies to the inconveniences stated above are the pfBufferAddChild(),
pfBufferRemoveChild(), and pfBufferClone() functions. The first two functions are
identical to their non-buffer counterparts pfAddChild() and pfRemoveChild() except
the buffer versions do not happen immediately. Other functions, pfBufferAdd(),
pfBufferInsert(), pfBufferReplace(), and pfBufferRemove(), perform the
buffer-oriented delayed-action versions of the corresponding non-buffer pfList
functions. In all cases the add, insert, replace, or removal request is placed on a list in the
current pfBuffer and is processed later at pfMergeBuffer() time.
The pfBufferClone() function supports the notion of maintaining a library of common
objects like trees or houses in a special library pfBuffer. The main database process then
clones objects from the library pfBuffer into the database pfBuffer, possibly using the
pfFlatten() function for improved rendering performance. pfBufferClone() is identical
to pfClone() except the buffer version requires that the source pfBuffer be specified and
that all cloned nodes have scope in the source pfBuffer.
pfAsyncDelete
We have discussed how to create subgraphs for database paging: create and select a
current pfBuffer, create nodes and build the subgraph, call pfBufferAddChild() and
finally pfMergeBuffer() to incorporate the subgraph into the application’s scene. This
section describes how to use the function pfAsyncDelete() to free the memory of old,
unwanted subgraphs.
164
007-1680-100
Successful Multiprocessing with OpenGL Performer
The pfDelete() function is the normal mechanism for deleting objects and freeing their
associated memory. However,the function pfDelete() can be a very expensive since it
must traverse, unreference, and register a deletion request for every OpenGL Performer
object it encounters which has a 0 reference count. The function pfAsyncDelete() used in
conjunction with a forked DBASE process moves the burden of deletion to the
asynchronous database process so that all rendering and intersection pipelines are not
adversely affected.
The pfAsyncDelete() function may be called from any process and places an
asynchronous deletion request on a global list that is processed later by the DBASE stage
when its trigger function pfDBase() is called. A major difference from pfDelete() is that
pfAsyncDelete() does not immediately check the reference count of the object to be
deleted and, so, does not return a value indicating whether the deletion was successful.
At this time there is no way of querying the result of a pfAsyncDelete() request so care
should be taken that the object to be deleted has no reference counts or memory leaks will
result.
Placing Multiple OpenGL Performer Processes on a Single CPU
When placing multiple OpenGL Performer processes on the same CPU, some
combinations of processes and priorities may have an effect on the APP process timing
even if the APP process runs on its own separate CPU. This happens because the APP
process often waits on other processes for completion of various tasks. If these other
processes share a CPU with high-priority processes, they may take a long time to finish
their task and release the APP process.
An application can request that OpenGL Performer upgrade the priority of processes
when the APP process waits on them by calling pfProcessPriorityUpgrade(). The APP
process upgrades the other process’ priority before it starts waiting for it, and the other
process resumes its previous priority as soon as it releases the APP process. In this way,
the original settings of priorities is maintained, except when the APP process waits for
another process. OpenGL Performer uses the priority 87 as the default priority for
upgrading processes. This priority is the default because it is close to the highest priority
that any application-level process should ever have (89). The application may change this
priority by using pfProcessHighestPriority().
The priority-upgrade mode is turned off by default. An OpenGL Performer application
that does not try to place multiple processes on the same processor or a non-realtime
application does not have to set this flag.
007-1680-100
165
5: Frame and Load Control
Rules for Invoking Functions While Multiprocessing
There are some restrictions on which functions can be called from an OpenGL Performer
process while multiple processes are running. Some specialized processes (such as the
process handling the draw stage) can call only a few specific OpenGL Performer
functions and cannot call any other kinds of functions. This section lists general and
specific rules concerning function invocation in the various OpenGL Performer and user
processes.
In this section, the phrase “the draw process” refers to whichever process is handling the
draw stage, regardless of whether that process is also handling other stages. Similarly,
“the cull process” and “the application process” refer to the processes handling the cull
and application stages, respectively.
This is a general list of the kinds of routines you can call from each process:
application
Configuration routines, creation and deletion routines, set and get
routines, and trigger routines such as pfAppFrame(), pfSync(), and
pfFrame()
database
Creation and deletion routines, set and get routines, pfDBase(), and
pfMergeBuffer()
cull
pfCull(), pfCullPath(), OpenGL Performer graphics routines
draw
pfClearChan(), pfDraw(), pfDrawChanStats(), OpenGL Performer
graphics routines, and graphics library routines
More specific elaborations:
•
166
You should call configuration routines only from the application process, and only
after pfInit() and before pfConfig(). pfInit() must be the first OpenGL Performer
call, except for those routines that configure shared memory (see “Memory
Allocation” in Chapter 18). Configuration routines do not take effect until
pfConfig() is called. These are the configuration routines:
–
pfMultipipe()
–
pfMultiprocess()
–
pfMultithread()
–
pfHyperpipe()
007-1680-100
Successful Multiprocessing with OpenGL Performer
•
You should call creation routines, such as pfNewChan(), pfNewScene(), and
pfAllocIsectData(), only in the application process after calling pfConfig() or in a
process that has an active pfBuffer. There is no restriction on creating libpr objects
like pfGeoSets and pfTextures.
•
The pfDelete() function should only be called from the application or database
processes while pfAsyncDelete() may be called from any process.
•
Read-only routines—that is, the pfGet*() functions—can be called from any
OpenGL Performer process. However, if a forked draw process queries a pfNode,
the data returned will not be frame-accurate. (See “Multiprocessing and Memory”
on page 168.)
•
Write routines—functions that set parameters—should be called only from the
application process or a process with an active pfBuffer. It is possible to call a write
routine from the cull process, but it is not recommended since any modifications to
the database will not be visible to the application process if it is separate from the
cull (as when using PFMP_APP_CULLDRAW or PFMP_APP_CULL_DRAW).
However, for transient modifications like custom level-of-detail switching, it is
reasonable for the cull process to modify the database. The draw process should
never modify any pfNode.
•
OpenGL Performer graphics routines should be called only from the cull or draw
processes. These routines may modify the hardware graphics state. They are the
routines that can be captured by an open pfDispList. (See “Display Lists” in
Chapter 12.) If invoked in the cull process, these routines are captured by an
internal pfDispList and later invoked in the draw process; but if they are invoked in
the draw process, they immediately affect the current window. These graphics
routines can be roughly partitioned into those that do the following:
•
007-1680-100
–
Apply a graphics entity: pfApplyMtl(), pfApplyTex(), and pfLightOn().
–
Enable or disable a graphics mode: pfEnable() and pfDisable().
–
Set or a modify graphics state: pfTransparency(), pfPushState(), and
pfMultMatrix().
–
Draw geometry or modify the screen: pfDrawGSet(), pfDrawString(), and
pfClear().
Graphics library routines should be called only from the draw process. Since there is
no open display list to capture these commands, an open window is required to
accept them.
167
5: Frame and Load Control
•
“Trigger” routines should be called only from the appropriate processes (see
Table 5-4).
Table 5-4
Trigger Routines and Associated Processes
Trigger Routine
Process/Context
pfAppFrame()
pfSync()
pfFrame()
APP/main loop
pfPassChanData()
pfPassIsectData()
APP/main loop
pfApp()
APP/channel APP callback
pfCull()
pfCullPath()
CULL/channel CULL callback
pfDraw()
pfDrawBin()
DRAW/channel DRAW callback
pfNodeIsectSegs()
pfChanNodeIsectSegs()
ISECT/callback or APP/main loop
pfDBase()
DBASE/callback
•
User-spawned processes created with sproc() can trigger parallel intersection
traversals through multiple calls to pfNodeIsectSegs() and
pfChanNodeIsectSegs().
•
Functions pfApp(), pfCull(), pfDraw(), and pfDBase() are only called from within
the corresponding callback specified by pfChanTravFunc() or pfDBaseFunc().
Multiprocessing and Memory
In OpenGL Performer, as is often true of multiprocessing systems, memory management
is the most difficult aspect of multiprocessing. Most data management problems in an
OpenGL Performer application can be partitioned into three categories:
•
168
Memory visibility. OpenGL Performer uses fork(), which—unlike sproc()—
generates processes that do not share the same address space. The processes also
cannot share global variables that are modified after the fork() call. After calling
fork(), processes must communicate through explicit shared memory.
007-1680-100
Successful Multiprocessing with OpenGL Performer
•
Memory exclusion. If multiple processes read or write the same chunk of data at the
same time, consequences can be dire. For example, one process might read the data
while in an inconsistent state and end up dumping core while dereferencing a
NULL pointer.
•
Memory synchronization. OpenGL Performer is configured as a pipeline where
different processes are working on different frames at the same time. This pipelined
nature is illustrated in Figure 5-6 on page 162, which shows that, for instance, in the
PFMP_APP_CULL_DRAW configuration the application process is working on
frame n while the draw process is working on frame n–2. If, in this case, if we have
only a single memory location representing the viewpoint, then it is possible for the
application to set the viewpoint to that of frame n and the draw process to
incorrectly use that same viewpoint for frame n–2. Properly synchronized data is
called frame accurate.
Fortunately, OpenGL Performer transparently solves all of the problems just described
for most OpenGL Performer data structures and also provides powerful tools and
mechanisms that the application can use to manage its own memory.
Shared Memory and pfInit()
The pfInit() function creates a shared memory arena that is shared by all processes
spawned by OpenGL Performer and all user processes that are spawned from any
OpenGL Performer process. A handle to this arena is returned by pfGetSharedArena()
and should be used as the arena argument to routines that create data that must be visible
to all processes. Routines that accept an arena argument are the pfNew*() routines found
in the libpr library and the OpenGL Performer memory allocator, pfMalloc(). In
practice, it is usually safest to create libpr objects like pfGeoSets and pfMaterials in
shared memory. libpf objects like pfNodes are always created in shared memory.
Allocating shared memory does not by itself solve the memory visibility problem
discussed above. You must also make sure that the pointer that references the memory is
visible to all processes. OpenGL Performer objects, once incorporated into the database
through routines like pfAddGSet(), pfAddChild(), and pfChanScene(), automatically
ensure that the object pointers are visible to all OpenGL Performer processes.
However, pointers to application data must be explicitly shared. A common way of
doing this is to allocate the shared memory after pfInit() but before pfConfig() and to
reference the memory with a global pointer. Since the pointer is set before pfConfig()
forks any processes, these processes will all share the pointer’s value and can thereby
007-1680-100
169
5: Frame and Load Control
access the same shared memory region. However, if this pointer value changes in a
process, its value will not change in any other process, since forked processes do not
share the same address space.
Even with data visible to all processes, data exclusion is still a problem. The usual
solution is to use hardware spin locks so that a process can lock the data segment while
reading or writing data. If all processes must acquire the lock before accessing the data,
then a process is guaranteed that no other processes will be accessing the data at the same
time. All processes must adhere to this locking protocol, however, or exclusion is not
guaranteed.
In addition to a shared memory arena, pfInit() creates a semaphore arena whose handle
is returned by pfGetSemaArena(). Locks can be allocated from this semaphore arena by
usnewlock() and can be set and unset by ussetlock() and usunsetlock(), respectively.
pfDataPools
The pfDataPools—named shared memory arenas with named allocation blocks—
provide a complete solution to the memory visibility and memory exclusion problems,
thereby obviating the need to set global pointers between pfInit() and pfConfig(). For
more information about pfDataPools, see the pfDataPools man page.
Passthrough Data
The techniques discussed thus far do not solve the memory synchronization problem.
OpenGL Performer’s libpf library provides a solution in the form of passthrough data.
When using pipelined multiprocessing, data must be passed through the processing
pipeline so that data modifications reach the appropriate pipeline stage at the
appropriate time.
Passthrough data is implemented by allocating a data buffer for each stage in the
processing pipeline. Then, at well-defined points in time, the passthrough data is copied
from its buffer into the next buffer along the pipeline. This copying guarantees memory
exclusion, but you should minimize the amount of passthrough data to reduce the time
spent copying.
Allocate a passthrough data buffer for the rendering pipeline using pfAllocChanData();
for data to be passed down the intersection pipeline, call pfAllocIsectData(). Data
returned from pfAllocChanData() is passed to the channel cull and draw callbacks that
170
007-1680-100
CULL Process Optimizations
are set by pfChanTravFunc(). Data returned from pfAllocIsectData() is passed to the
intersection callback specified by pfIsectFunc().
Passthrough data is not automatically passed through the processing pipeline. You must
first call pfPassChanData() or pfPassIsectData() to indicate that the data should be
copied downstream. This requirement allows you to copy only when necessary—if your
data has not changed in a given frame, simply do not call a pfPass*() routine, and you
will avoid the copy overhead. When you do call a pfPass*() routine, the data is not
immediately copied but is delayed until the next call to pfFrame(). The data is then
copied into internal OpenGL Performer memory and you are free to modify your
passthrough data segment for the next frame.
Modifications to all libpf objects—such as pfNodes and pfChannels—are
automatically passed through the processing pipeline, so frame-accurate behavior is
guaranteed for these objects. However, in order to save substantial amounts of memory,
libpr objects such as pfGeoSets and pfGeoStates do not have frame-accurate behavior;
modifications to such objects are immediately visible to all processes. If you want
frame-accurate modifications to libpr objects you must use the passthrough data
mechanism, use a frame-accurate pfSwitch to select among multiple copies of the objects
you want to change, or use the pfCycleBuffer memory type.
CULL Process Optimizations
The OpenGL Performer CULL process traverses a scene graph and culls out invisible
geometry. Its result is a list (pfDispList) of visible pfGeoSets. The OpenGL Performer
CULL process treats pfGeoSets as rendering atoms: It does not break them into their
visible and invisible parts. If the bounding box of a pfGeoSet intersects the viewing
frustum, OpenGL Performer draws the entire pfGeoSet even if only one of its triangles is
visible. Figure 5-7 demonstrates this problem using a triangle strip.
007-1680-100
171
5: Frame and Load Control
Figure 5-7
Loose Culling of pfGeosets
The figure shows a triangle strip starting inside the viewing frustum, leaving the viewing
frustum, and then returning into the viewing frustum. Only the shaded triangles of the
strip are visible, but OpenGL Performer renders the entire strip. In this figure,
OpenGL Performer sends five superfluous vertices to the graphics pipe.
This problem is important in applications with one of the following bottlenecks:
•
Geometry processing
Applications that render large numbers of relatively small triangles—for example,
CAD visualization or detailed terrain visualization.
•
Host-Pipe interface bandwidth
Applications that saturate the interface between the host CPU and the graphics pipe
either by rendering too many triangles or by downloading too many texture maps
each frame.
This problem is not important in raster-limited applications that render very large
triangles (in screen space). These application saturate the raster portion of the graphics
pipe but leave the geometry portion idle. Therefore, speeding up the geometry portion
of the graphic pipe does not speed up the overall application frame rate.
172
007-1680-100
CULL Process Optimizations
Cull Sidekick Processes
You can overcome the loose-culling problem by allocating extra processes for cleaning up
the pfGeoSet lists that the CULL processes produce. These extra processes are called
CULL sidekicks. By default, a CULL sidekick process checks all the primitives in all
pfGeoSets on the CULL output. It replaces original pfGeoSets with temporary pfGeoSets
and populates the temporary pfGeoSets with the visible parts of the original pfGeoSets.
By default, CULL sidekick processes test each primitive twice for the following:
•
For frustum visibility
A primitive outside the viewing frustum will be omitted from the temporary
pfGeoSet.
•
For backface culling
A primitive facing away from the viewer will be omitted from the temporary
pfGeoSet. This test is especially powerful when rendering enclosed objects (for
example—vehicles, houses, or machine parts) because about half of the triangles in
such models face away from the viewer. This test is skipped when a pfGeoSet is
drawn without backface testing.
CULL sidekick processes run side-by-side with their CULL process. They do not interact
with the CULL process during its frame, but they merely patch the visible pfGeoSet list
as the CULL process populates it. This means that configuring CULL sidekick processes
does not add any latency to the application.
Figure 5-8 shows how CULL_SIDEKICK optimizes visible pfGeoSet lists while CULL is
writing them. The figure shows three CULL_SIDEKICK processes working on the visible
pfGeoSet list that a CULL process produces. Visible pfGeoSet#1 is replaced by
Temporary pfGeoSet#1. Visible pfGeoSet#2 contains no visible primitives and is skipped
entirely.
007-1680-100
173
5: Frame and Load Control
Temporary
pfGeoSet #1
CULL_SIDEKICK
Skipped
Test and replace visible pfGeoSets
Visible
pfGeoSet
list
Write visible pfGeoSets
visible
pfGeoSet #1
Figure 5-8
visible
pfGeoSet #2
visible
pfGeoSet #3
CULL
CULL_SIDEKICK Processing
Configuring CULL_SIDEKICK Processes
Each CULL process can have multiple CULL_SIDEKICK processes. You can use the
pfMultithread() call to specify the number of CULL_SIDEKICK processes for each CULL
process. The collection of CULL_SIDEKICK processes configured for each CULL process
traverses in a round-robin manner the pfGeoSet list that the CULL process produces. The
more CULL_SIDEKICK processes (each assigned to a separate CPU), the faster they
process the pfGeoSet list that the CULL process produces.
CULL Sidekick Optimization Mask
Using the function pfMultithreadParami() and the parameter PFSK_OPTIMIZATION,
an application can specify a bit-wise OR of the constants PFSK_BACKFACE_CULL and
PFSK_FRUSTUM_CULL. Specifying the PFSK_BACKFACE_CULL flag instructs
CULL_SIDEKICK to run a backface test on each primitive and to remove backfacing
primitives. This mode is aware of the pfGeoState setting for each pfGeoSet and correctly
ignores pfGeoSets that do not require this test. Specifying the PFSK_FRUSTUM_CULL
flag instructs CULL_SIDEKICK to run a frustum test on each primitive and to remove
primitives outside the viewing frustum. Both of these tests break triangle strips, line
strips, and triangle fans if portions of these are invisible.
174
007-1680-100
CULL Process Optimizations
Note: It is safe to change the CULL_SIDEKICK optimization mask on the fly.
CULL Sidekick Synchronization Policy
Since traversing the visible pfGeoSet list that CULL produces may take longer than a
single frame, you can specify a policy for the behavior of CULL_SIDEKICK processes.
Using the function pfMultithreadParami() and the parameter PFSK_POLICY, you can
specify one of three options:
•
PFSK_CULL_DONE
All CULL_SIDEKICK processes stop processing pfGeoSet lists as soon as their
CULL process finishes its frame. This means that the CULL_SIDEKICK process is
likely to skip the optimization of many pfGeoSets on the visible pfGeoSet list.
•
PFSK_CULL_FRAME_DONE
All CULL_SIDEKICK processes continue processing until the end of the expected
CULL frame time. If the CULL process finishes its frame early in the
PFSK_CULL_DONE mode, the CULL_SIDEKICK processes cannot use the
remainder of the time to complete their own processing. The
PFSK_CULL_FRAME_DONE mode allows the CULL_SIDEKICK processes to use
all of the available frame time for processing. Use the parameter
PFSK_SAFETY_MARGIN to specify a floating number of seconds. This sets a
margin before the end of the frame where CULL_SIDEKICK stops processing. This
is a safety measure. If CULL_SIDEKICK does not complete early enough, it can
make CULL miss its frame. The default value is 1.0 millisecond. The more sensitive
to frame drops your application is, the larger this margin should be.
•
PFSK_CULL_SIDEKICK_DONE
All CULL_SIDEKICK processes finish processing all the visible pfGeoSet lists that
the CULL process produces. If this takes longer than the desired CULL frame rate,
the CULL process waits for its CULL_SIDEKICK helpers and may miss a frame.
Note: It is safe to change the CULL_SIDEKICK synchronization policy on the fly.
007-1680-100
175
5: Frame and Load Control
CULL Sidekick User Functions
Use the function pfMultithreadParami() with parameters PFSK_USER_FUNC and
PFSK_USER_FUNC_DATA to register a callback function for the CULL_SIDEKICK
pfGeoSet optimization. When specified, a CULL_SIDEKICK calls the callback function
instead of running the default optimization. The CULL_SIDEKICK provides the callback
function with a target pfGeoSet. The callback function can clone the target pfGeoSet,
modify the cloned pfGeoSet, and return it as a replacement for the target pfGeoSet.
The callback function should return a pfGeoSet pointer. It can return one of the following
values:
•
The original pfGeoSet pointer
CULL_SIDEKICK does not optimize this pfGeoSet and leaves it on the visible
pfGeoSet list.
•
A new pfGeoSet pointer
CULL_SIDEKICK replaces the pfGeoSet in the visible pfGeoSet list with the
returned value.
•
A NULL pointer
CULL_SIDEKICK removes this pfGeoSet from the visible pfGeoSet list.
The callback function receives as a parameter a pointer to a pfDispListOptimizer class.
The callback function can use this pointer in order to do the following:
176
•
Retrieve the projection/modelview matrix that will be loaded when this pfGeoSet is
rendered.
•
Allocate temporary pfGeoSets.
•
Allocate temporary memory buffers.
•
Clone a pfGeoSet onto a temporary pfGeoSet.
•
Invoke the default optimization on a pfGeoSet.
•
Get a pointer to the pfChannel in which this pfGeoSet was found visible.
•
Get the number of CULL_SIDEKICK processes working for the CULL process and
get the index of the calling CULL_SIDEKICK process.
•
Get the optimization mask of this CULL_SIDEKICK process.
007-1680-100
CULL Process Optimizations
The following is a sample callback function. This function clones the incoming pfGeoSet,
jitters all its coordinates by a random amount, and replaces all its colors by random
colors:
pfGeoSet *
userFunction(pfGeoSet *gset, pfDispListOptimizer *op, void *userData)
{
pfGeoSet
*new_gset;
ushort
*ilist;
int
*len;
float
*c;
float
*v;
int
numVerts, numPrims, numColors;
int
i;
/* Modify geosets with line-strip/tri-strip primitives only. */
/* When not modifying a pfGeoSet, return its original pointer. */
if ((pfGetGSetPrimType(gset) != PFGS_LINESTRIPS) &&
(pfGetGSetPrimType(gset) != PFGS_TRISTRIPS))
return (gset);
/* Clone geoset. We can modify the cloned geoset because it */
/* is temporary for this CULL process for this frame. */
new_gset = pfDLOptimizerCloneGSet(op, gset,
PFSK_COORD3 |PFSK_NORMAL3 |
PFSK_TEXCOORD2 |PFSK_ATTR_LENGTHS);
/* Get pointers to cloned geoset attributes */
pfGetGSetAttrLists(new_gset, PFGS_COLOR4, &c, &ilist);
if (ilist) return gset; /* ignore indexed gsets */
pfGetGSetAttrLists(new_gset, PFGS_COORD3, &v, &ilist);
if (ilist) return gset; /* ignore indexed gsets */
len = pfGetGSetPrimLengths(new_gset);
numPrims = pfGetGSetNumPrims(new_gset);
/* Count how many vertex entries in the COORD3 attribute. */
numVerts = 0;
for (i = 0 ; i < numPrims ; i ++)
numVerts += len[i];
/* Count how many color entries in the COLOR4 attribute. */
switch (pfGetGSetAttrBind(gset, PFGS_COLOR4))
{
case PFGS_PER_VERTEX:
numColors = numVerts;
007-1680-100
177
5: Frame and Load Control
break;
case PFGS_PER_PRIM:
numColors = numPrims;
break;
case PFGS_OVERALL:
numColors = 1;
break;
case PFGS_OFF:
numColors = 0;
break;
}
/* Pick a random color for each color entry in the cloned */
/* color attribute array. */
for (i = 0 ; i < numColors ; i ++)
{
*(c++) = getRand(); *(c++) = getRand(); *(c++) = getRand();
*(c++) = 1.0;
}
/* Pick a random perturbation for each coordinate */
for (i = 0 ; i < numVerts ; i ++)
{
*(v++) += vertex_jitter_amount * getRand();
*(v++) += vertex_jitter_amount * getRand();
*(v++) += vertex_jitter_amount * getRand();
}
/* Send new geoset for default frustum/backface culling. */
return pfDLOptimizerOptimize(op, new_gset);
}
Modifying Attributes of Cloned pfGeoSets
When cloning a pfGeoSet from within a CULL_SIDEKICK callback function, you may
wish to modify the pointers to the attribute arrays of the cloned pfGeoSet. Cloned
pfGeoSets are temporary and do not require reference counting. Use the following quick
methods on the pfGeoSet in order to manipulate its attributes:
•
pfQuickCopyGSet()
Copies the contents of one pfGeoSet onto another with no reference count
considerations.
•
pfGSetQuickAttr()
Sets an attribute of a pfGeoSet.
178
007-1680-100
CULL Process Optimizations
•
pfGSetQuickMultiAttr()
Sets a multi-value attribute of a pfGeoSet (for example, multitexture)
•
pfGSetQuickPrimLengths()
Sets the primitive length array of a pfGeoSet.
•
pfQuickResetGSet()
Sets all attribute arrays to NULL. No reference counting.
Note: If you wish to replace the attribute binding of cloned pfGeoSet attributes, you
must use the standard pfGeoSet API (as opposed to the quick API). Changing anything
other than the pointers to attribute arrays requires internal pfGeoSet state changes and,
therefore, cannot happen through the quick API.
Marking pfGeoSets for Optimization
Use the function pfGSetOptimize() to mark any single pfGeoSet for optimization by the
CULL_SIDEKICK process. By default, all pfGeoSets under a pfGeode node undergo
optimization. All pfGeoSetCBs are not optimized by default but can be optimized using
this function. No pfGeoSet under a pfBillboard node is ever optimized (regardless of the
optimization flag setting).
007-1680-100
179
Chapter 6
6. Creating Visual Effects
This chapter describes how to use environmental, atmospheric, lighting, and other visual
effects to enhance the realism of your application. The following sections appear:
•
“Using pfEarthSky” on page 181
•
“Atmospheric Effects” on page 182
•
“Patchy Fog and Layered Fog” on page 186
•
“Real-Time Shadows” on page 198
•
“Image-Based Rendering” on page 204
Using pfEarthSky
A pfEarthSky is a special set of functions that clears a pfChannel’s viewport efficiently
and implements various atmospheric effects. A pfEarthSky is attached to a pfChannel
with pfChanESky(). Several pfEarthSky definitions can be created, but only one can be
in effect for any given channel at a time.
A pfEarthSky can be used to draw a sky and horizon, to draw sky, horizon, and ground,
or just to clear the entire screen to a specific color and depth. The colors of the sky,
horizon, and ground can be changed in real time to simulate a specific time of day. At the
horizon boundary, the ground and sky share a common color, so that there is a smooth
transition from sky to horizon color. The width of the horizon band can be defined in
degrees.
A pfChannel’s earth-sky model is automatically drawn by OpenGL Performer before the
scene is drawn unless the pfChannel has a draw callback set with pfChanTravFunc(). In
this case it is the application’s responsibility to clear the viewport. Within the callback
pfClearChan() draws the channel’s pfEarthSky.
007-1680-100
181
6: Creating Visual Effects
Example 6-1 shows how to set up a pfEarthSky().
Example 6-1
How to Configure a pfEarthSky
pfEarthSky *esky;
pfChannel *chan;
sky = pfNewESky();
pfESkyMode(esky, PFES_BUFFER_CLEAR, PFES_SKY_GRND);
pfESkyAttr(esky, PFES_GRND_HT, -1.0f);
pfESkyColor(esky, PFES_GRND_FAR, 0.3f, 0.1f, 0.0f, 1.0f);
pfESkyColor(esky, PFES_GRND_NEAR, 0.5f, 0.3f, 0.1f,1.0f);
pfChanESky(chan, esky);
Atmospheric Effects
The complexities of atmospheric effects on visibility are approximated within OpenGL
Performer using a multiple-layer sky model, set up as part of the pfEarthSky function. In
this design, individual layers are used to represent the effects of ground fog, clear sky,
and clouds. Figure 6-1 shows the identity and arrangement of these layers.
182
007-1680-100
Atmospheric Effects
General
visibility
Upper
transition
zone
Clouds
Lower
transition
zone
General
visibility
Groung fog
Figure 6-1
Layered Atmosphere Model
The lowest layer consists of ground fog, extending from the ground up to a user-selected
altitude. The fog thins out with increasing altitude, disappearing entirely at the bottom
of the general visibility layer. This layer extends from the top of the ground fog layer to
the bottom of the cloud layer’s lower transition zone, if such a zone exists. The transition
007-1680-100
183
6: Creating Visual Effects
zone provides a smooth transition between general visibility and the cloud layer. (If there
is no cloud layer, then general visibility extends upward forever.) The cloud layer is
defined as an opaque region of near-zero visibility; you can set its upper and lower
boundaries. You can also place another transition zone above the cloud layer to make the
clouds gradually thin out into clear air.
Set up the atmospheric simulation with the commands listed in Table 6-1
Table 6-1
pfEarthSky Functions
Function
Action
pfNewESky()
Create a pfEarthSky.
pfESkyMode()
Set the render mode.
pfESkyAttr()
Set the attributes of the earth and sky models.
pfESkyColor()
Set the colors for earth and sky and clear.
pfESkyFog()
Set the fog functions.
You can set any pfEarthSky attribute, mode, or color in real time. Selecting the active
pfFog definition can also be done in real time. However, changing the parameters of a
pfFog once they are set is not advised when in multiprocessing mode.
The default characteristics of a pfEarthSky are listed in Table 6-2.
Table 6-2
184
pfEarthSky Attributes
Attribute
Default
Clear method
PFES_FAST (full screen clear)
Clear color
0.0 0.0 0.0
Sky top color
0.0 0.0 0.44
Sky bottom color
0.0 0.4 0.7
Ground near color
0.5 0.3 0.0
Ground far color
0.4 0.2 0.0
Horizon color
0.8 0.8 1.0
007-1680-100
Atmospheric Effects
Table 6-2
pfEarthSky Attributes (continued)
Attribute
Default
Ground fog
NULL (no fog)
General visibility
NULL (no fog)
Cloud top
20000.0
Cloud bottom
20000.0
Cloud bottom color
0.8 0.8 0.8
Cloud top color
0.8 0.8 0.8
Transition zone bottom
15000.0
Transition zone top
25000.0
Ground height
0
Horizon angle
10 degrees
By default, an earth-sky model is not drawn. Instead, the channel is simply cleared to
black and the Z-buffer is set to its maximum value. This default action also disables all
other atmospheric attributes. To enable atmospheric effects, select PFES_SKY,
PFES_SKY_GRND, or PFES_SKY_CLEAR when turning on the earth-sky model.
Clouds are disabled when the cloud top is less than or equal to the cloud bottom. Cloud
transition zones are disabled when clouds are disabled.
Fog is enabled when either the general or ground fog is set to a valid pfFog. If ground fog
is not enabled, no ground fog layer will be present and fog will be used to support
general visibility. Setting a fog attribute to NULL disables it. See “Atmospheric Effects”
on page 182 for further information on fog parameters and operation.
The earth-sky model is an attribute of the channel and thus accesses information about
the viewer’s position, current field of view, and other pertinent information directly from
pfChannel. To set the pfEarthSky in a channel, use pfChanESky().
007-1680-100
185
6: Creating Visual Effects
Patchy Fog and Layered Fog
A pfVolFog is a class that uses a multi-pass algorithm to draw the scene with a fog that
has different densities at different locations. It extends the basic layered fog provided by
pfEarthSky and introduces a new type of fog: a patchy fog. A patchy fog has a constant
density in a given area. The boundaries of this area can be defined by an arbitrary
three-dimensional object or by a set of objects.
A layered fog changes only with elevation; its density and color is uniform at a given
height. It is defined by a set of elevation points, each specifying a fog density and,
optionally, also a fog color at the point’s elevation. The density and the color between two
neighboring points is linearly interpolated.
Figure 6-2 illustrates the basic difference between patchy fog and layered fog.
P1
P2
color 2
P3
P4
P5
node 1
color 1
P6
node 2
Layered fog
Patchy fog
Figure 6-2
Patchy Fog Versus Layered Fog
Compared to a layered fog in pfEarthSky, a layered fog in pfVolFog has distinct
advantages:
186
•
It can be specified by an arbitrary number of elevation points.
•
Each elevation point can have a different color associated with it.
•
A layered fog in pfVolFog is not dependent on an InfiniteReality-specific texgen. It
can also be drawn using only 2D textures to simulate the 3D texture. Thus, a layered
fog in pfVolFog can virtually be used on any machine.
007-1680-100
Patchy Fog and Layered Fog
Creating Layered Fog
A pfVolFog is not part of the scene graph; it is created separately by the application
process. Once created, elevation points of a layered fog can be specified by calling
pfVolFogAddPoint() or pfVolFogAddColoredPoint() repeatedly. The fog initialization
is completed by calling pfApplyVolFog().
Example 6-2
Fog initialization Using pfVolFogAddPoint()
pfVolFog *lfog;
lfog = pfNewVolFog(arena);
pfVolFogAddPoint(lfog, elev1, density1);
pfVolFogAddPoint(lfog, elev2, density2);
pfVolFogAddPoint(lfog, elev2, density2);
pfApplyVolFog(lfog);
Creating Patchy Fog
The boundary of a patchy fog is specified by pfVolFogAddNode(pfog,node),where node
contains the surfaces enclosing the foggy areas. It is possible to define several disjoint
areas in the same tree or by adding several different nodes. Note that each area has to be
completely enclosed, and the vertices of the surfaces have to be ordered so that the front
face of each surface faces outside the foggy area. The node has to be part of the scene
graph for the rendering to work properly.
Example 6-3
Specifying Patchy Fog Boundaries Using pfVolFogAddNode()
pfVolFog *pfog;
pfNode
*fogNode;
pfog = pfNewVolFog(arena);
fogNode = pfdLoadFile(filename);
pfVolFogAddNode(pfog, fogNode);
pfAddChild(scene, fogNode);
pfApplyVolFog(pfog);
Patchy and layered fog can be combined but only if layered fog has a uniform color; that
is, it is specified using pfVolFogAddPoint() only.
007-1680-100
187
6: Creating Visual Effects
Initializing a pfVolFog
The function pfApplyVolFog() initializes a pfVolFog. If at least two elevation points were
defined, it initializes data structures necessary for rendering of a layered fog, including
a 3D texture. Any control points defined afterward are ignored. If a node containing
patchy fog boundaries has been added prior to calling pfApplyVolFog(), a patchy fog is
initialized. Since function pfVolFogAddNode() only marks the parts of the scene graph
that specifies the texture, it is possible to add additional patchy fog nodes, even after
pfApplyVolFog() has been called.
Table 6-3 summarizes routines for initialization and drawing of a pfVolFog.
Table 6-3
188
pfVolFog Functions
Function
Action
pfNewVolFog()
Create a pfVolFog.
pfVolFogAddChannel()
Add a channel on which pfVolFog is used.
pfVolFogAddPoint()
Add a point specifying fog density at a certain elevation.
pfVolFogAddColoredPoint()
Add a point specifying fog density and color at a certain elevation.
pfVolFogAddNode()
Add a node defining the boundary of a patchy fog.
pfVolFogSetColor()
Set color of a layered fog or patchy fog.
pfVolFogSetDensity()
Set density of a patchy fog.
pfVolFogSetFlags()
Set binary flags.
pfVolFogSetVal()
Set a single attribute.
pfVolFogSetAttr()
Set an array of attributes.
pfApplyVolFog()
Initialize data structures necessary for rendering fog.
pfVolFogAddChannel()
Add a channel on which pfVolFog is used.
pfVolFogUpdateView()
Update the current view for all stored channels.
pfDrawVolFog()
Draw the scene with fog.
pfGetVolFogTexture()
Return the texture used by layered fog.
007-1680-100
Patchy Fog and Layered Fog
The attributes of a pfVolFog are listed in Table 6-4.
Table 6-4
pfVolFog Attributes
Attribute
Identifier
Default
Color
PFVFOG_COLOR
0.9, 0.9, 1
Density
PFVFOG_DENSITY
1.0
Density bias
PFVFOG_DENSITY_BIAS
0
Maximum distance
PFVFOG_MAX_DISTANCE
2000
Mode
PFVFOG_MODE
PFVFOG_LINEAR
Layered fog mode
PFVFOG_LAYERED_MODE
PFVFOG_LINEAR
Texture size
PFVFOG_3D_TEX_SIZE
64 x 64 x 64
Resolution
PFVFOG_RESOLUTION
0.2
Patchy fog mode
PFVFOG_PATCHY_MODE
PFVFOG_LINEAR
Texture bottom
PFVFOG_PATCHY_TEXTURE_BOTTOM
0.3
Texture top
PFVFOG_PATCHY_TEXTURE_TOP
0.1.5
PFVFOG_ROTATE_NODE
Identity
Attenuation scale
PFVFOG_LIGHT_SHAFT_ATTEN_SCALE
0.04
Attenuation shift
PFVFOG_LIGHT_SHAFT_ATTEN_TRANSLATE 6
Darken factor
PFVFOG_LIGHT_SHAFT_DARKEN_FACTOR
General
Layered fog
Patchy fog
Layered patchy fog
Rotation matrix
Light shafts
007-1680-100
0.3
189
6: Creating Visual Effects
The flags of a pfVolFog are listed in Table 6-5.
Table 6-5
pfVolFog Flags
Flag
Identifier
Default
Close surfaces
PFVFOG_FLAG_CLOSE_SURFACES
1
Use 2D texture
PFVFOG_FLAG_FORCE_2D_TEXTURE
0
Force patchy fog passes
PFVFOG_FLAG_FORCE_PATCHY_PASS
0
Self-shadowing
PFVFOG_FLAG_SELF_SHADOWING
0
Darken objects
PFVFOG_FLAG_DARKEN_OBJECTS
0
Filter color
PFVFOG_FLAG_FOG_FILTER
0
Faster patchy fog
PFVFOG_FLAG_FASTER_PATCHY_FOG
0
No object in fog
PFVFOG_FLAG_NO_OBJECT_IN_FOG
0
1D texture on surface
PFVFOG_FLAG_PATCHY_FOG_1DTEXTURE
0
Separate node bins
PFVFOG_FLAG_SEPARATE_NODE_BINS
0
Screen-bounding rectangle
PFVFOG_FLAG_SCREEN_BOUNDING_RECT
1
Draw nodes separately
PFVFOG_FLAG_DRAW_NODES_SEPARATELY
0
User-defined texture
PFVFOG_FLAG_USER_PATCHY_FOG_TEXTURE
0
Use cull programs
PFVFOG_FLAG_USE_CULL_PROGRAM
0
PFVFOG_FLAG_LAYERED_PATCHY_FOG
0
PFVFOG_FLAG_LIGHT_SHAFT
0
General
Layered fog
Patchy fog
Layered patchy fog
Use layered patchy fog
Light shafts
Light shaft
190
007-1680-100
Patchy Fog and Layered Fog
Updating the View
A pfVolFog needs information about the current eye position and view direction. Since
this information is not directly accessible in a draw process, it is necessary to call
pfVolFogAddChannel() for each channel at the beginning of the application. Whenever
the view changes, the application process has to call pfVolFogUpdateView(). See
programs in /usr/share/Performer/src/sample/apps/C/fogfly or
/usr/share/Performer/src/sample/apps/C++/volfog on IRIX and Linux or
%PFROOT%\Src\sample\apps\C\fogfly or
%PFROOT%\Src\sample\apps\C++\volfog on Microsoft Windows for an example.
If you do not update the view, the fog will not be rendered.
If the application changes the position of the patchy fog boundaries (for example, by
inserting a pfSCS, pfDCS, or pfFCS node above the fog node) or the orientation of the
whole scene with respect to the up vector (for example, the use of a trackball in Perfly),
the fog may not be drawn correctly.
Drawing a Scene with Fog
To draw the scene with a fog, the draw process has to call pfDrawVolFog() instead of
pfDraw(). This function takes care of drawing the whole scene graph with the specified
fog. Expect the draw time to increase because the scene is drawn twice (three times if both
patchy and layered fog are specified). In case of a patchy fog there may also be several
full-screen polygons being drawn. You can easily disable the fog by not calling
pfDrawVolFog().
Since boundaries of patchy fog are in the scene graph, do not use pfDraw() to draw the
scene without fog; instead, use pfDrawBin() with PFSORT_DEFAULT_BIN,
PFSORT_OPAQUE_BIN, and PFSORT_TRANSP_BIN.
A patchy fog needs as deep a color buffer as possible (optimally 12 bits per color
component) and a stencil buffer. Use at least a 4-bit stencil buffer (1-bit is sufficient only
for very simple fog objects). It may be necessary to modify your application so that it asks
for such a visual.
Deleting a pfVolFog
A pfVolFog can be deleted using pfDelete(). In case of a layered fog it is necessary to
delete the texture handle in a draw process. The texture is returned by
007-1680-100
191
6: Creating Visual Effects
pfGetVolFogTexture(). See the example in
/usr/share/Performer/src/sample/apps/C/fogfly on IRIX and Linux and in
%PFROOT%\Src\sample\apps\C\fogfly on Microsoft Windows.
Specifying Fog Parameters
This section describes how to manage the various parameters for both layered and
patchy fog.
Layered Fog
As mentioned earlier, a layered fog of a uniform color is specified by function
pfVolFogAddPoint(), which sets the fog density at a given elevation. The density is
scaled so that if the fog has a density of 1, the nearest object inside the fog that has full
fog color is at a distance equal to 1/10 of the diagonal of the scene bounding box. The
layered fog color is set by function pfVolFogSetColor() or by calling pfVolFogSetAttr()
with parameter PFVFOG_COLOR and a pointer to an array of three floats.
A layered fog of nonuniform color is specified by function pfVolFogAddColoredPoint(),
which sets the fog density and the fog color at a given elevation. The color set by
pfVolFogSetColor() is then ignored.
The layered fog mode is set by function pfVolFogSetVal() with parameter
PFVFOG_LAYERED_MODE and one of PFVFOG_LINEAR, PFVFOG_EXP, or
PFVFOG_EXP2.
It is also possible to set the mode both for a layered and patchy fog at once by using
parameter PFVFOG_MODE. The default mode is PFVFOG_LINEAR. The function of the
mode parameter is equivalent to the function of the fog mode parameter of the OpenGL
function glFog().
The size of a 3D texture used by a layered fog can be modified by calling
pfVolFogSetAttr() with parameter PFVFOG_3D_TEX_SIZE and an array of three integer
values. The default texture size is 64x64x64, but reasonable results can be achieved with
even smaller sizes. The sizes are automatically rounded up to the closest power of 2. The
second value should be equal to or greater than the third value. If 3D textures are not
supported, a set of 2D textures is used instead of a 3D texture (the number of 2D textures
is equal to the third dimension of the 3D texture). Every time the r coordinate changes
more than 0.1, a new texture is computed by interpolating between two neighboring
192
007-1680-100
Patchy Fog and Layered Fog
slices, and the texture is reloaded. The use of 2D textures can be forced by calling:
pfVolFogSetFlags() with flag PFVFOG_FLAG_FORCE_2D_TEXTURE set to 1.
Note: Once a layered fog is initialized by calling the pfApplyVolFog(), changing any of
the parameters described here will not affect rendering of the layered fog.
Patchy Fog
The density of a patchy fog is controlled by function pfVolFogSetDensity() or by using
pfVolFogSetVal() with parameter PFVFOG_FOG_DENSITY. As in the case of a layered
fog, the density of a patchy fog is scaled by 1/10 of the diagonal of the scene bounding
box.
You can specify an additional density value that is added to every pixel inside or behind
a patchy fog boundary using the function pfVolFogSetVal() with parameter
PFVFOG_FOG_DENSITY_BIAS. This value makes a patchy fog appear denser but it
may create unrealistically sharp boundaries.
The patchy fog color is set by function pfVolFogSetColor() or by calling
pfVolFogSetAttr() with parameter PFVFOG_COLOR and a pointer to an array of three
floats. If the blend_color extension is not available, patchy fog will be white.
The patchy fog mode is set by function pfVolFogSetVal() with parameter
PFVFOG_PATCHY_MODE and one of PFVFOG_LINEAR, PFVFOG_EXP, or
PFVFOG_EXP2.
It is also possible to set the mode both for a patchy and layered fog at once by using
parameter PFVFOG_MODE. The default mode is PFVFOG_LINEAR.
Note: The parameters of a patchy fog can be modified at any time and they will affect
the rendering of the subsequent frame.
Advanced Features of Layered Fog and Patchy Fog
This section describes the following topics:
•
007-1680-100
“Enabling Self-Shadowing of a Layered Fog and Scene Darkening”
193
6: Creating Visual Effects
•
“Animating Patchy Fog”
•
“Selecting a Different Type of Patchy Fog Algorithm”
•
“Simulating Self-Shadowing in Patchy Fog”
•
“Layered Patchy Fog”
•
“Light Shafts”
The example in /usr/share/Performer/src/sample/C++/volfog on IRIX and
Linux and in %PFROOT%\Src\sample\C++\volfog on Microsoft Windows illustrates
the use of all these advanced features.
Enabling Self-Shadowing of a Layered Fog and Scene Darkening
A layered fog can be self-shadowed—that is, the lower parts of a dense fog appear
darker. Self-shadowing is enabled by setting the flag
PFVFOG_FLAG_SELF_SHADOWING to 1. The fog mode should be set to
PFVFOG_EXP.
When the fog has different colors at different elevations and the flag
PFVFOG_FLAG_FOG_FILTER is set to 1, a secondary scattering is approximated. In this
case, the color of a higher layer may affect the color of a lower layer.
If the flag PFVFOG_FLAG_DARKEN_OBJECTS is set, even the objects below a dense fog
become darker. The light is assumed to come from the top.
Animating Patchy Fog
A patchy fog can be animated by modifying the geometry of the fog nodes. When
changing the content of geosets specifying the fog boundary, make sure that the geosets
are fluxed and that the bounding box of each geoset is updated. In addition, function
pfVolFogAddNode() has to be called every time the fog bounding box changes.
Selecting a Different Type of Patchy Fog Algorithm
It is possible to use a different algorithm for rendering patchy fog that can handle
semi-transparent surfaces better. To use this algorithm, set the flag
PFVFOG_FASTER_PATCHY_FOG to 1. Some advanced features of patchy fog described
in the following subsections are supported only in one of the two algorithms. In such
cases, this limitation is noted.
194
007-1680-100
Patchy Fog and Layered Fog
Simulating Self-Shadowing in Patchy Fog
If the flag PFVFOG_FASTER_PATCHY_FOG is set to 1, the algorithm also allows the
color of the patchy fog boundary to be modified using a texture. Either a built-in 1D
texture expressing the attenuation between two elevations is used or you can provide a
1D or a 3D texture for each volume object. This can be used to simulate self-shadowing
of dense gases, such as clouds.
The built-in 1D texture is enabled by setting the flag
PFVFOG_FLAG_PATCHY_FOG_1DTEXTURE. The texture is mapped to the range of
elevations between the bottom and top of the fog bounding box. The texture value at the
bottom (default of 0.3) can be modified by calling pfVolFogSetVal() with parameter
PFVFOG_PATCHY_TEXTURE_BOTTOM and the value at the top (default of 1.5) using
parameter PFVFOG_PATCHY_TEXTURE_TOP.
To use a different scale for objects of different sizes, you must specify the fog objects
separately. When the flag PFVFOG_FLAG_SEPARATE_NODE_BINS is set, all calls to
pfVolFogAddNode() define fog nodes that are drawn separately, and the predefined
texture is scaled according to the bounding box of each node.
If both the flag PFVFOG_FLAG_PATCHY_FOG_1DTEXTURE and the flag
PFVFOG_FLAG_USER_PATCHY_FOG_TEXTURE are set, textures associated with the
fog nodes are used to modify the surface color of a patchy fog.
To avoid artifacts on overlapping colored patchy fog objects the flag
PFVFOG_FLAG_DRAW_NODES_SEPARATELY forces the algorithm to be applied to
each node separately in the back-to-front order with respect to the viewpoint. Currently,
this mode does not work well when scene objects intersect fog objects.
Layered Patchy Fog
If the flag PFVFOG_FLAG_LAYERED_PATCHY_FOG is set, the layered fog is used to
define the density of a patchy fog. The layered fog is then present only in areas enclosed
by the patchy fog boundaries. Since layered fog is computed for the whole scene, it is
important to set fog parameter PFVFOG_MAX_DISTANCE to a value that corresponds
to the size of the patchy fog area (for example, a diameter of its bounding sphere). Use
function pfVolFogSetVal() to modify the maximum distance parameter.
Layered patchy fog nodes can be moved and rotated by specifying a matrix for each fog
node, identified by its index (the order in which nodes were specified). The function
pfVolFogSetAttr() with three parameters specified can be used for this purpose. The first
007-1680-100
195
6: Creating Visual Effects
parameter is PFVFOG_ROTATE_NODE, the second parameter specifies the node index,
and the last one is a pointer to a pfMatrix.
Light Shafts
Light shafts are a special application of a layered patchy fog. The fog boundary specifies
a cone of light with decreasing intensity (density) along the cone axis. Additional
rendering passes darken the objects outside the cone of light and lighten the objects
inside the light shaft based on their distance from the light. To enable these additional
passes, set flag PFVFOG_FLAG_LIGHT_SHAFT to 1. To ensure that these passes are
applied even if the light shaft is not in the field of view, you must also set flag
PFVFOG_FLAG_FORCE_PATCHY_PASS to 1.
To control the additional passes, the parameter
PFVFOG_LIGHT_SHAFT_DARKEN_FACTOR (set using pfVolFogSetAttr()) can
change the factor by which all objects outside the light shaft are darkened. The default
value is 0.3.
Parameters PFVFOG_LIGHT_SHAFT_ATTEN_SCALE and
PFVFOG_LIGHT_SHAFT_ATTEN_TRANSLATE set the translate and scaling of a
built-in, one-dimensional texture that is used to reduce the color of objects lit by the light.
Set the translate to a small value—for example, 10 to 20% of the shaft length—and the
scale to the inverse of the shaft length.
Performance Considerations and Limitations
The quality and speed of patchy fog rendering can be controlled by calling
pfVolFogSetVal() with the parameter PFVFOG_RESOLUTION. The resolution is a value
between 0 and 1. Higher values will reduce banding and speed up the drawing. On the
other hand, high values may cause corruption in areas of many overlapping fog surfaces.
The default value is 0.2, but you may use values higher than that if your fog boundaries
do not overlap much.
The following are other performance considerations:
•
196
The multipass algorithms used for rendering layered and patchy fog may produce
incorrect results if the scene graph contains polygons that have equal depth values.
To avoid such problems, a stencil buffer is used during rendering of the second
pass. You can disable this function by setting the flag
PFVFOG_FLAG_CLOSE_SURFACES to 0.
007-1680-100
Patchy Fog and Layered Fog
007-1680-100
•
By default, the multipass algorithm is applied only when boundaries of a patchy
fog are visible. This may cause undesirable changes of semi-transparent edges of
scene objects when fog objects move into or away from the view. To force the use of
the multipass algorithm, set the flag PFVFOG_FLAG_FORCE_PATCHY_PASS to 1.
•
Cull programs (see “Cull Programs” in Chapter 4) can speed up rendering of patchy
fog because in some draw passes only the part of the scene intersecting the fog
boundary is rendered. To enable cull programs, set the flag
PFVFOG_FLAG_USE_CULL_PROGRAM to 1.
•
A layered fog is faster to render than a patchy fog; use a layered fog instead of a
patchy fog whenever possible. Rendering of both types of fog together is even
slower; so, you may try to define only one type.
•
Changing the fog mode does not affect the rendering speed in the case of a layered
fog but rendering of a patchy fog is slower for fog modes PFVFOG_EXP and
PFVFOG_EXP2. If you prefer using non-linear modes, try to use them only for
layered fog and not for patchy fog.
•
You can speed up drawing of a patchy fog by reducing the size of the fog
boundaries. In case of several disjoint fog areas, the size of a bounding box
containing all boundaries will affect the draw time and quality. Try to avoid
defining a patchy fog in two opposite parts of your scene. Try also to increase the
value of resolution (if there are not too many overlapping fog boundaries) or reduce
the patchy fog density.
•
If there is a lot of banding visible in the fog, try to choose a visual with as many bits
per color component as possible. Keep in mind that a patchy fog needs a stencil
buffer. You can also try to apply all techniques mentioned in the previous item—
reducing the size of patchy fog boundaries, increasing resolution, or decreasing
density.
•
If a patchy fog looks incorrect (the fog appears outside the specified boundaries)
make sure that the vertices of the fog boundaries are specified in the correct order so
that front faces always face outside the foggy area.
•
If you see a darker band in a layered fog at eye level, make sure the texture size is
set so that the second value is equal to or greater than the third value.
•
Since light shafts are using a combination of layered and patchy fog and the density
is decreasing to 0 at the end of the light cone, the quality of results is very sensitive
to the depth of color buffers. 12-bit visuals are required and the light shaft should
not be too large. Also, ensure that PFVFOG_MAX_DISTANCE is set as small as
possible.
197
6: Creating Visual Effects
OpenGL Performer has the following limitations in regards to fog management:
Layered fog
•
The values of a layered fog are determined at each vertex and interpolated across a
polygon. Consequently, an object located on top of a large ground polygon may be
fogged a bit more or less than the part of the polygon just under the object.
•
A layered fog works fast with a 3D texture. Reloading of 2D textures during the
animation can be slow.
Patchy fog
•
The method does not work well for semitransparent surfaces. If your scene contains
objects that are semitransparent or that have semitransparent edges, (for example,
tree billboards or mountains in Performer Town), these objects or edges may be cut
or may be fogged more than the neighboring pixels. Even if a semitransparent edge
of a billboard is outside the fog, it will not be smooth.
•
A layered patchy fog is extremely sensitive to the size of the fog area and the
density of the layered fog. Specifically, the fog values accumulated along an
arbitrary line crossing the bounding box of the fog area should not reach 1.
•
A patchy fog needs a stencil buffer and the deepest color buffers possible.The
rendering quality on a visual with less than 12 bits per color component is low
unless the fogged area is very small compared to the size of the whole scene.
•
If the blend_color extension is not available, the patchy fog color will be white.
Real-Time Shadows
You can create real-time shadows using the class pfShadow. You specify a set of light
sources and a set of objects that cast shadows on all other objects in the scene. The class
manages the drawing and renders shadows for each combination of a shadow caster and
a light source. Shadows are rendered by projecting the objects as seen from the light
source into a texture and projecting the texture onto a scene. To avoid computing the
texture for each frame, a set of textures is precomputed at the first frame, then for each
frame the best representative is chosen and warped to approximate the correct shadow.
The following sections further describe real-time shadows:
198
•
“Creating a pfShadow”
•
“Drawing a Scene with Shadows”
007-1680-100
Real-Time Shadows
•
“Specifying Shadow Parameters”
•
“Assigning Data with Directions”
•
“Limitations of Real-Time Shadows”
Creating a pfShadow
A pfShadow is not part of the scene graph; it is created separately by the application
process. Once the pfShadow is created, you can specify the number of shadow casters by
calling function pfShadowNumCasters() and then set each caster using the function
pfShadowShadowCasters(). Each shadow caster is specified by a scene graph node and
a matrix that contains the transformation of the node with respect to the scene graph root.
Shadow casters are indexed from 0 to the number of casters minus 1.
Similarly, the number of light sources is set by function pfShadowNumSources(). A light
source is defined by its position or direction, set by pfShadowSourcePos() or
pfShadowLight().
A pfShadow needs information about the current eye position and view direction. Since
this information is not directly accessible in a draw process, it is necessary to call
pfShadowAddChannel() for each channel at the beginning of the application. Whenever
the view changes, the application process has to call pfShadowUpdateView(). Even if
the view does not change, this function must be called at least once in single-process
mode or as many times as the number of buffers in a pfFlux in multiprocess mode.
Without updating the view, the shadow is not rendered correctly.
The class initialization is completed by calling the function pfShadowApply() as shown
in the following creation example:
pfShadow *shd = pfNewShadow();
pfShadowNumCasters(shd, 2);
pfShadowShadowCaster(shd, 0, node1, matrix1);
pfShadowShadowCaster(shd, 1, node2, matrix2);
pfShadowNumSources(shd, 1);
pfShadowSourcePos(shd, 0, x1, y1, z1, w1);
pfShadowAddChannel(channel);
pfShadowApply(shd);
007-1680-100
199
6: Creating Visual Effects
Table 6-6 summarizes the functions for the initialization and drawing of a pfShadow.
Table 6-6
200
pfShadow Functions
Function
Action
pfNewShadow()
Create a pfShadow.
pfShadowNumCasters()
Set number of shadow casters.
pfShadowShadowCaster()
Set a shadow caster and its rotation matrix.
pfShadowAdjustCasterCenter()
Specify the translation of caster's center.
pfShadowNumSources()
Set number of light sources.
pfShadowSourcePos()
Specify light source position.
pfShadowLight()
Specify light source.
pfShadowAmbientFactor()
Set ambient factor.
pfShadowShadowTexture()
Set a user-defined shadow texture for a given caster
and light source.
pfShadowTextureBlendFunc()
Set a function used when blending closest shadows.
pfShadowAddChannel()
Add a channel on which pfShadow is used.
pfShadowUpdateView()
Update the current view for all stored channels.
pfShadowUpdateCaster()
Update rotation matrix of a caster.
pfShadowFlags()
Set binary flags.
pfShadowVal()
Set a single attribute.
pfGetShadowDirData()
Get a pfDirData associated with the pfShadow.
pfShadowApply()
Initialize a pfShadow.
pfShadowDraw()
Draw the scene and shadows.
007-1680-100
Real-Time Shadows
The attributes of a pfShadow are listed in Table 6-7.
Table 6-7
pfShadow Attributes
Attribute
Identifier
Default
Size of shadow texture
PFSHD_PARAM_TEXTURE_SIZE
512 x 512
Number of shadow textures
PFSHD_PARAM_NUM_TEXTURES
1
There is only one pfShadow flag, PFSHD_BLEND_TEXTURES. This blend-textures flag
has a default of 0.
Drawing a Scene with Shadows
To draw a scene with real-time shadows, the draw process has to call the draw function
provided by the pfShadow class: pfShadowDraw(). Before the first frame is rendered, all
required shadow textures are precomputed. A warning is printed if the window size is
smaller than the texture dimensions. Ensure that the window is not obscured; otherwise,
the textures will not be correct.
By default, only the closest shadow texture is selected for any direction and it is skewed
so that it approximates the correct shadow. Optionally, the flag
PFSHD_BLEND_TEXTURES can be set using the function pfShadowFlags(). In this case,
the two closest textures are selected and blended together, resulting in smoother
transitions. Also, instead of a linear blend between the textures, you can define a blend
function, mapping values 0–1 to the interval 0–1. The blend function can be set using the
function pfShadowTextureBlendFunc().
Every time the caster changes its position or orientation with respect to the light source,
it is necessary to update its matrix using pfShadowUpdateCaster() (the caster is
identified by its index). When the caster's matrix changes, the shadow of the caster
changes as well. In this case, the set of precomputed shadow textures is searched to find
the one or two closest representatives.
007-1680-100
201
6: Creating Visual Effects
Specifying Shadow Parameters
The shadow texture is used to darken the scene pixels when the texture texel is set to 1.
The amount by which the scene pixel is darkened can be set by the function
pfShadowAmbientFactor(). The default value is 0.6
As the caster is projected into a shadow texture, the center of the projection corresponds
with the center of the bounding box of the caster's node. When the shadow texture is
skewed to approximate shadows from a slightly different direction, it is best if the center
of the projection corresponds with the center of the object. The bounding box center may
not coincide with the center of the object (in the case of some long protruding parts) and
you can use the function pfShadowAdjustCasterCenter() to shift the bounding box
center toward the center of the object.
For each combination of a shadow caster and a light source, it is possible to specify the
number of shadow textures used, their sizes, and a set of directions for which the textures
are precomputed. The number of textures and their sizes can be set by the function
pfShadowVal(), where the first parameter is PFSHD_PARAM_TEXTURE_SIZE or
PFSHD_PARAM_NUM_TEXTURES.
The set of directions can be controlled by using the function pfGetShadowDirData() to
get the pointer to the corresponding pfDirData, a class that stores data associated with a
set of directions. Then you can either select the default mode or specify the directions
directly. See following section “Assigning Data with Directions” for more details. By
default, there is one texture of size 512 x 512 and the direction corresponds to the light
direction (or a vector from a point light source to the object's center). If there are more
textures, the original light direction is rotated around a horizontal direction, assuming
that the object will primarily keep its horizontal position (for example, a helicopter or a
plane).
A sample implementation of shadows is in the file
perf/samples/pguide/libpf/C++/shadowsNew.
Assigning Data with Directions
The pfDirData class is used to store directional data—that is, data that depend on
direction. A pfDirData stores an array of directions and an array of (void *) pointers
representing the data associated with each direction.
202
007-1680-100
Real-Time Shadows
The directions and data can be set using the function pfDirDataData(). Optionally, you
can set only the directions using the function pfDirDataDirections() in the case that the
associated data are defined later or generated internally by another OpenGL Performer
class (such as pfShadow).
You can also generate directions automatically using the function
pfDirDataGenerateDirections(). The first parameter defines one of the default sets of
directions and the second parameter is used to specify additional values. At present only
type PFDD_2D_ROTATE_AROUND_UP is supported, in which case the second
parameter points to a 3D vector that is rotated around the up vector, creating a number
of directions.
The data can be queried using the pfDirDataFindData() or pfDirDataFindData2()
function. In the first case, the function finds the closest direction to the direction specified
as the first parameter, copies it to the second parameter, and returns the pointer to the
data associated with it. The input direction has to be normalized. The second function
finds the two closest directions to the specified direction. It copies the two directions to
the second parameter (which should point to an array of two vectors). The two pointers
to the data associated with the two directions are copied to the array of two (void *)
pointers specified as the third parameter. In addition, two weights associated with each
direction are copied to the array of two floats. These weights are determined based on the
distance of the end point of the input direction and each of the two closest directions.
Limitations of Real-Time Shadows
The following are limitations of real-time shadows in OpenGL Performer:
•
When projecting a caster into a shadow texture, pfSwitch children are selected
according to switch value. In the case of pfLOD, the finest level is chosen. Also,
pfSequences are ignored—which can be useful in the case of helicopter rotors, for
example.
•
The pfShadow class uses cull programs to cull out geometry that is not affected by
the shadow to make the multipass drawing more efficient. At present, though, the
cull program used by the pfShadow class overwrites any other cull program you
specify.
Note: Ensure that you do not overwrite TravMode in your application by setting it
to PFCULL_ALL. The mode is set by pfShadow when pfShadowApply() is called.
007-1680-100
203
6: Creating Visual Effects
•
When projecting a caster into a shadow texture, pfSwitch and pfLOD may not be
handled properly. Also, pfSequences are ignored—which can be useful in case of
helicopter rotors, for example.
Image-Based Rendering
The image-based rendering approach is used for very complex objects. Such an object is
represented by a set of images taken from many directions around it. When the object is
rendered for each view direction, several closest views are blended together.
In OpenGL Performer, you can use the pfIBRnode class to represent complex objects.
Unlike a pfBillboard, a parent class of pfIBRnode, the texture on pfGeoSets of a
pfIBRnode is not static, but it changes based on the view direction for each pfGeoSet.
The following sections further describe image-based rendering:
•
“Creating a pfIBRnode” on page 204
•
“Creating a pfIBRnode Using a Proxy” on page 205
•
“Creating a pfIBRtexture” on page 206
•
“Parameters Controlling Drawing of a pfIBRnode” on page 208
•
“The Simplify Application” on page 209
•
“Creating Images of an Object with makeProxyImages” on page 215
•
“Creating Images of an Object with makeIBRimages” on page 219
•
“Limitations” on page 220
Creating a pfIBRnode
A pfIBRnode is a child class of pfBillboard. You create a pfIRRnode in a fashion similar
to that of a pfBillboard. Compared to a pfBillboard, a pfIBRnode has two additional
parameters: a pfIBRtexture and an array of angles defining the initial rotation of the
objects.
Each pfIBRnode has associated with it a single pfIBRtexture, which stores a set of images
of the complex object as viewed from different directions. Each pfGeoSet is then rendered
204
007-1680-100
Image-Based Rendering
with a texture representing the view of the object from the given direction. A
pfIBRtexture is specified using the function pfIBRnodeIBRtexture().
Using the function pfIBRnodeAngles(), you control the initial orientation of the complex
object by specifying the rotation from the horizontal and vertical planes for each
pfGeoSet. These angles are very useful in case of trees, for example, because you can use
a different vertical angle for each instance of the tree. The trees then appear different,
although they all use the same pfIBRtexture. The first value is ignored in the case that
only one ring of views around the object is used.
You must set up a pfIBRnode so that the pfIBRtexture applied to it can modify properly
the image at each frame. You do so in the following manner:
1.
Set the texture of the pfGeoState associated with each pfGeoSet of the pfIBRnode to
the texture returned by the function pfGetIBRtextureDefaultTexture().
2. If the pfIBRtexture has the flag PFIBR_USE_REG_COMBINERS set, enable
multitexturing and specify texture coordinates for additional texture units.
3. If the pfIBRtexture has the flag PFIBR_3D_VIEWS enabled, set the billboard rotation
(PFBB_ROT) to PFBB_AXIAL_ROT.
On IRIX and Linux, see the example in the following file:
/usr/share/Performer/src/sample/pguide/C++/IBRnode.C
On Microsoft Windows, see the example in the following file:
%PFROOT%\Src\sample\pguide\C++\IBRnode.C
Creating a pfIBRnode Using a Proxy
By default, it is assumed that the geosets of the pfIBRnode specify rectangles that are
always facing the viewer (like billboards). This approach is very fast but it requires a
large number of views to limit the artifacts due to the differences between the
neighboring views.
To reduce the number of views required to obtain a reasonable image of the complex
object from any direction, we can use a shape that approximates the surface of the
complex object instead of a billboard. This shape is called a proxy. The closer the proxy
is to the original surface, the fewer views of the objects are required. Optimally, you
create a proxy that contains a relatively small number of primitives and that is very close
007-1680-100
205
6: Creating Visual Effects
to the original surface. The proxy can be created using the new tool Simply. See section
“The Simplify Application” on page 209 for the details.
Compared to default mapping of views on a billboard there are only minor changes.
Instead of a billboard, the node's geosets contain the proxy geometry. The pfIBRtexture
associated with the node has the flag PFIBR_USE_PROXY set. There is an array of texture
coordinates indexed by the view index and the geoset index. These texture coordinates
can be defined and queried by pfIBRnodeProxyTexCoords() and
pfGetIBRnodeProxyTexCoords(). Note that it is more efficient to store the proxy in one
geoset.
Optionally, it is possible to specify different geosets for each view (if the
PFIBR_NEAREST flag is set in the pfIBRtexture assigned to the pfIBRnode) or for each
group of views if the views are blended. In this case, you must set the flag
PFIBRN_VARY_PROXY_GEOSETS using pfIBRnodeFlags(). This can be useful for
removing the invisible parts of the proxy (invisible from the range of views in the group)
or for sorting the proxy triangles to avoid artifacts when edges of the proxy textures are
transparent. The array of texture coordinates is then organized as follows:
•
The first index is the view index or the group index (if the views are blended).
•
The second index is the geoset index multiplied by the number of views in a group
(1 for the nearest view).
•
The coordinates are grouped by geosets.
Thus, there are texture coordinates for the geoset 0 for all views in the group, then for
geoset 1, and so on.
The geosets are organized as follows: if the proxy has n geosets and there are v views or
groups of views, the pfIBRnode has n*v geosets, and each group of n geosets belongs to
one view.
To create views of a complex object from various directions and to compute the texture
coordinates of its proxy, you can use the makeProxyImages tool described in section
“Creating Images of an Object with makeProxyImages” on page 215.
Creating a pfIBRtexture
A pfIBRtexture stores a set of images of a complex object as viewed from different
directions. The directions are specified using pfIBRtextureIBRdirections(). Internally,
206
007-1680-100
Image-Based Rendering
pfIBRtexture uses pfDirData to store the views. A pfDirData determines the type of view
distribution. It could be a set of views around the object with all views perpendicular to
the vertical axis, or the views can be from a set of rings and each ring contains an array
of evenly spaced views that have the same angle from the horizontal plane. Otherwise,
the views are assumed to be uniformly or randomly distributed around the sphere of
directions. You must specify the directions before the images are set.
Once you specify the directions, you set the images using pfIBRtextureIBRtextures().
The parameters are an array of pointers to the textures containing the views and the
number of the textures in this array.
If views are organized in rings, you can load the images directly from a set of files using
pfIBRtextureLoadIBRtexture() without the need to specify the directions first. The
parameter format specifies the path where the images are stored as well as how they are
indexed—for example, images/view%03d.rgb. The other two parameters specify the
number of images and the increment between two loaded images. The increment
specification is useful when the texture memory is limited; for instance, specifying
step=2 causes every second image to be skipped. Optionally, you can specify the views
using the function pfIBRtextureIBRtextures(). The parameters are an array of pointers
to the textures containing the views and the number of the textures in this array.
If the views are organized in rings, the textures, by default, represent views around the
object, all perpendicular to the vertical axis. In this case, specified textures form a single
ring of views that are evenly spaced. If the flag PFIBR_3D_VIEWS is specified by the
function pfIBRtextureFlags(), the textures form a set of rings. Each ring contains an array
of evenly spaced views that have the same angle from the horizontal plane.
If the flag PFIBR_3D_VIEWS is not set, both functions pfIBRtextureLoadIBRtexture()
and pfIBRtextureIBRtextures() will set one ring with the specified number of textures
and a horizontal angle of 0. If the flag PFIBR_3D_VIEWS is set, the class checks whether
a file info is present in the image directory. If it is, the information about rings is loaded
from that file. The file contains two values on each line: the horizontal angle and the
number of textures at each ring. If the file is not present in the image directory, you must
specify the rings before the images are loaded by calling the functions
pfIBRtextureNumRings() and pfIBRtextureRing(). Rings are indexed from 0 and
should be ordered by the horizontal angle, with the lowest angle at index 0. Each ring can
have a different number of textures associated with it.
When 3D views are used, the image files read by function
pfIBRtextureLoadIBRtexture() should be indexed by the ring index and the index of the
007-1680-100
207
6: Creating Visual Effects
image in a given ring. Specify the format string in the manner shown in the following
example:
images/view%02d_%03d.rgb
If you specify the textures using the function pfIBRtextureIBRtextures(), the texture
pointers are all stored in a single array, starting with textures of the first ring, followed
by textures of the second ring, and so on.
It is assumed that the views in each ring are uniformly spaced and they are ordered
clockwise with respect to the vertical axis. If the views are ordered in the opposite
direction, use the function pfIBRtextureDirection() to set the direction to –1.
When using pfIBRnodes and pfIBRtextures in Perfly, you need an alpha buffer. If the
pfIBRnode is rendered as an opaque rectangle, try the command-line parameter –9, in
which case Perfly requests a visual with an alpha buffer.
For more details about associating a pfIBRtexture with a pfIBRnode, see the pfIBRnode
man page and the following program:
/usr/share/Performer/src/sample/pguide/C++/IBRnode (IRIX and Linux)
%PFROOT%\Src\sample\pguide\C++\IBRnode (Microsoft Windows)
Parameters Controlling Drawing of a pfIBRnode
At present, the pfIBRtexture class is used only by the pfIBRnode class. The pfIBRtexture
class provides a draw function for pfGeoSets that belong to the pfIBRnode, but the draw
process is transparent to you. You can control the drawing by setting flags using the
function pfIBRtextureFlags(). If the flag PFIBR_NEAREST is set, the closest view from
the closest ring is selected and applied as a texture of the pfGeoSet. This approach is fast
on all platforms, but it results in visible jumps when the texture is changed. Thus, by
default, the flag PFIBR_NEAREST is not set and the two or, in case of 3D views, four
closest views are blended together. If the graphics hardware supports register combiners,
flags PFIBR_USE_REG_COMBINERS and PFIBR_USE_2D_TEXTURES are
automatically set by the class constructor and blending of textures can be done in one
pass.
The flag PFIBR_USE_PROXY is used when the views are mapped on an approximation
of the complex object (a proxy) and a different draw function is applied. You can read
more about proxies in section “Creating a pfIBRnode Using a Proxy” on page 205.
208
007-1680-100
Image-Based Rendering
By default on IRIX, the flag PFIBR_USE_2D_TEXTURES is not set and a 3D texture is
used for fast blending between the two closest views. To avoid flickering when the object
is viewed from a distance, additional 3D textures are used to store additional mipmap
levels. This feature is available on machines with multisampling only (InfiniteReality
systems). To disable the mipmapping, set flag PFIBR_MIPMAP_3DTEXTURES to zero.
In case of several rings, the nearest ring is selected and the views inside this ring are
blended using the 3D texture. 3D texture is not compatible with other distributions of the
views. Hence, in this case, ensure that you set flag PFIBR_USE_2D_TEXTURES.
The Simplify Application
The Simplify application is an interactive tool that is used to simplify a complex object.
It has the following two main functions:
•
Create a regular simplification of an object
•
Create a proxy of an object
In a regular simplification of an object, the resulting geometry does not cross the inner
and outer boundaries of the original object. The distance of these boundaries from the
original object controls the coarseness of the resulting geometry. All vertex parameters,
such as the normal or texture coordinates, are preserved. A simplified version of the
object can be used to create a pfLOD node (see section “pfLOD Nodes” on page 70).
A proxy is a simplified version of the object where the original object is fully inside the
proxy. This property is important because the proxy is used in image-based rendering
where the images of a complex object from various directions are projected onto the
proxy. In this way, it is possible to render a very complex object using a simplified version
(a proxy) and store the surface detail, including the associated lighting, in multiple
textures. See section “Creating Images of an Object with makeProxyImages” on page 215
for the process of making the textures that are projected on the proxy.
The Simplify Graphical User Interface (GUI)
The Simplify application is based on the Perfly application and they share many
command-line parameters and key commands (see the man page for perfly). The
syntax for the command-line invocation is as follows:
simplify [ perfly-options ] infile outfile [ simplification-options ]
007-1680-100
209
6: Creating Visual Effects
You can get the list of the simplication options by running simplify with no option or
with only the option –h.
When you start the Simplify application, the menu is similar to that of Perfly. There is an
additional pane of buttons and sliders, called the Simplify pane, which can be enabled
and disabled using the Simplify pane button. Figure 6-3 shows the Simplify pane, which
is enabled by default. Most of the buttons and sliders on the Simplify pane have
command-line equivalents.
Figure 6-3
210
The Default Simplify Pane
007-1680-100
Image-Based Rendering
Making a Proxy of an Object
Computing a proxy with Simplify requires two basic decisions:
•
Where to position the initial proxy and an outer boundary for the original object
•
What algorithm to use for creating the initial proxy and the outer boundary
Since these decisions may be difficult to make in an analytical fashion initially, the
Simplify GUI allows you to make some guesses and refine them in an iterative fashion.
The following procedure for making a proxy assumes that you have invoked the
Simplify application using the default simplification options.
1.
Ensure that the Simplify into proxy button is selected (the default).
2. Specify the initial distance of the proxy from the object and an outer boundary.
Use the sliders Initial distance and Outer boundary to do this. Distances are
specified as a percentage of the object diameter (more precisely, the diameter of the
object's bounding sphere). Initially, you might want to use the defaults, 2% for
Initial distance and 5% for Outer boundary.
3. Select the algorithm for creating the initial proxy.
Simplify provides two algorithms: the marching cubes algorithm and the
deplace-along-normals algorithm. The first button on the Simplify pane is the
Do marching cubes button, which is selected by default. If the Do marching cubes
button is not selected, Simplify uses the deplace-along-normals algorithm.
The marching cubes algorithm creates an isosurface at a certain distance (slider
Iso distance) from the original object. The isosurface is later moved to the distance
of the outer boundary (slider Outer boundary) and a copy of the isosurface is
moved to the distance of the initial proxy (slider Initial distance).
The marching cubes algorithm has the following additional controls:
•
Grid Size X slider
•
Grid Size Y slider
•
Grid Size Z slider
•
Iso distance slider
With these controls, you can set the grid size at each axis and the distance of the
isosurface from the object (using the slider Iso distance). The finer the grid, the
longer the algorithm takes and the more complex the initial proxy. On the other
hand, if the grid is too coarse, many details may be missed.
007-1680-100
211
6: Creating Visual Effects
In general, the algorithm does not work very well if the desired isosurface distance
is too small compared to the size of a grid voxel. For this reason, it is possible to
specify the isosurface distance separately from the outer boundary distance and the
initial proxy distance. Often it is possible to specify the isosurface distance large
enough so that the isosurface does not miss any part of the object and then move it
closer as needed. It is also possible to preview the isosurface by clicking the button
Get isosurface while the button Show boundary is selected.
If you select the deplace-along-normals algorithm, the outer boundary and the
initial proxy are created by displacing the original surface along its normals. This
approach works better in the case where distances are very small. Unfortunately,
some areas of the object may not be simplified. For example, if two parts of the
object are touching, displacing along the normals will create a self-intersecting
boundary that will not allow any room for simplification in the area of intersection.
With the deplace-along-normals algorithm, the grid is used to accelerate the
intersection test of the simplified proxy with the boundary surfaces. Thus, do not
reduce the grid resolution too much.
4. Click the Run simplify proxy button to start the simplification.
The simplification algorithm starts by moving the isosurface or the original surface
to create the outer boundary and the initial proxy. The initial proxy is simplified by
removing vertices and edges as long as the surface is within the surfaces defined by
the object and the outer boundary. At the end, the vertices of the proxy are moved as
close to the original object as possible.
After completing the computation, the proxy is saved in the file specified on the
command line.
The simplification algorithm can be stopped or paused by clicking the
Stop simplify or Pause button, respectively. When the algorithm is paused, it is
possible to save the current proxy by clicking the Save mesh button. The file name
contains the index of the current step so that several meshes can be output during
the simplification.
Simplifying an Object
The procedure for a regular simplification is very similar to the procedure for making a
proxy, as described in the preceding section. In contrast to making a proxy, however,
Simplify uses two boundary surfaces, an outer boundary (set by the slider
Outer boundary) and an inner boundary (set by the slider Inner boundary) to create a
regular simplication.
212
007-1680-100
Image-Based Rendering
To simplify an object requires two basic decisions:
•
Where to place an outer boundary and inner boundary
•
What algorithm to use for creating the boundaries
The following procedure for making a proxy assumes that you have invoked the
Simplify application using the default simplification options.
1.
Ensure that the Simplify into proxy button is not selected.
This is not the default. Figure 6-4 shows the resulting Simplify pane.
Figure 6-4
007-1680-100
The Simplify Pane for Simplifying an Object
213
6: Creating Visual Effects
2. Specify inner and outer boundaries.
Use the sliders Inner boundary and Outer boundary to do this. Distances are
specified as a percentage of the object diameter (more precisely, the diameter of the
object's bounding sphere). Initially, you might want to use the defaults, 2.5% for
Inner boundary and 5% for Outer boundary.
3. Select the algorithm for creating the boundaries.
Simplify provides two algorithms: the marching cubes algorithm and the
deplace-along-normals algorithm. The first button on the Simplify pane is the
Do marching cubes button, which is selected by default. If the Do marching cubes
button is not selected, Simplify uses the deplace-along-normals algorithm. See the
preceding section for a description of the algorithms.
If you select the marching cubes algorithm, the distance of both boundaries from the
original surface is the same (in absolute value) and it is controlled by the slider Iso
distance. As in the case of making a proxy, the isosurface can be previewed by
clicking the Get isosurface button.
If you select the deplace-along-normals algorithm, the boundaries are created by
moving the original surface along its normals to distances specified by the sliders
Outer boundary and Inner boundary. Note that the distance for the inner boundary
is specified as a negative number.
4. Click the Run simplify button to start the simplication.
The computation can be paused or stopped by clicking the Pause or Stop simplify
button, respectively. When the algorithm is paused, it is possible to save the
intermediate result by clicking the Save mesh button or to toggle the visibility of the
boundary by clicking the Show boundary button. After the simplification is
finished you can display the original object by clicking the Restore object button.
You can restart the algorithm without restoring the original object.
As you may realize, this procedure could be used to create an object proxy if you select
the displace-along-normals algorithm and the inner boundary is set to zero. The result
may be different, though, because the algorithm is trying to preserve seams between
pfGeoSets with different pfGeoStates; the seam preservation is not necessary for the
proxy.
214
007-1680-100
Image-Based Rendering
Creating Images of an Object with makeProxyImages
You can use the program makeProxyImages to create images (views) of the specified
object from a set of directions. Since the images are being projected on a proxy, a
simplification of the original object, additional processing may be required to add views
of parts of the proxy that are partially or fully obstructed by other parts. These additional
texture pieces are important because as the proxy is rotated away from the view at which
the texture was computed, some parts of the proxy that were not directly visible from the
view may become visible. Thus, each image consists of the view of the object and a
collection of texture pieces for obstructed parts of the proxy.
It is necessary to store texture coordinates for each proxy triangle so that the texture
pieces are correctly mapped. Consequently, the program makeProxyImages outputs
not only textures storing the views but also a pfIBRnode that contains the texture
coordinates and the proxy geometry. You can create the proxy of an object using the
program Simplify.
Command-Line Options for makeProxyImages
The input to the makeProxyImages program is the file containing the original complex
object. Table 6-8 and the following sections describe other key, command-line options:
•
“Packing Additional Textures Pieces” on page 216
•
“Fine Tuning Texture Rendering” on page 217
•
“Potential Problems” on page 218
Table 6-8
007-1680-100
Key Command-Line Options of makeProxyImages
Command
Option
Description
–pf
Specifies the file containing the proxy.
–f
Specifies the files where the images are stored. A view number and the extension
rgb is added automatically.
–pfb
Specifies the file where the resulting pfIBRnode is stored.
–W
Specifies the size of the texture (–W xsize ysize). It is important to specify the size.
–o
Specifies the oversampling factor. Specify this option when the hardware does not
support antialiasing.
215
6: Creating Visual Effects
Table 6-8
Command
Option
Key Command-Line Options of makeProxyImages (continued)
Description
–rv
Specifies the text file with ring information to determine the view directions. Each
line of the ring file contains two values: the angle from the horizontal plane and how
many views are created for that angle.
–n
Specifies that only views around the object are used.
–nv
Specifies that uniformly distributed 3D views are used.
–sk
Enables skipping a certain number of views in each ring.
–s
Scales up the object. By default, the program uses orthographic projection. The
center of the projection is the center of the bounding sphere around the object and
the object is scaled so that the bounding sphere fits the window. If the bounding
sphere is too large you may try to upscale the object using the –s option to make
better use of the texture. You can use perspective projection by defining the distance
of the camera from the center of the bounding sphere. Unless there are reasons for
doing otherwise, use the orthographic projection.
–l
Specifies non-default lighting. In image-based rendering the lighting is captured in
the textures. Thus, it is important you specify the lights in the same way as in your
scene. By default, the default Perfly lighting is selected. You can specify your own
lights using the –l option:
–l posx posy posz posw r g b
You can use multiple –l specifications to define multiple lights.
To obtain the full set of options, run the program makeProxyImages without any
parameters.
Packing Additional Textures Pieces
By default, the program makeProxyImages renders only the view of the object without
the extra texture pieces for obstructed triangles of the proxy. To enable this feature you
have to add the option –ev. The process has two steps. First, the number and size of
texture pieces is determined and a packing algorithm determines their position around
the primary view. Second, for each view the texture pieces are rendered in place.
The packing algorithm operates on the pixel level and there are several options that affect
its speed and the quality of the results. To speed up the packing algorithm, you can
216
007-1680-100
Image-Based Rendering
downsample the textures before packing using the option –evd. The drawback is that
there may be more wasted space between texture pieces. You can also reduce the number
of neighboring pixels the texture packing algorithm checks when finding the optimal
place for texture pieces by using the option –evp. In general, the texture pieces are not
aligned with their neighboring pieces. Thus, when the view texture is mipmapped, the
gaps between the textures may become visible. For this purpose, you can add the option
–evmp to set the number of mipmapping levels that will not have cracks. Each edge of
the texture piece that is not a silhouette edge is extended to contain more pixels from
neighboring triangles. Setting the value too high may cause the packing algorithm to fail.
If the packing algorithm fails to place the texture pieces around the primary view, the
object is scaled down a little (for the given view) and the algorithm is restarted. This
process repeats until all the texture pieces fit.
Similarly, as obstructed triangles may come into full view, backfacing triangles may
become visible as the proxy is rotated away from the view. Thus, it is possible to add
texture pieces for backfacing triangles into the view texture using the option –bf. Not all
backfacing triangles are added but only those that may be visible from neighboring
views. Since additional texture pieces that are used for backfacing triangles of the proxy
can be found in neighboring views, it is advantageous to combine several views into a
single texture. This reduces the number of texture pieces packed into a texture for one
view. You can use the option –tm to control this. Do not exceed 2Kx2K when combining
several views into one texture.
Fine Tuning Texture Rendering
When rendering additional texture pieces, you can control how far before and after the
proxy triangle the clip planes are being set. This option affects triangles around the
silhouette of the object. This is view-dependent: for each view, there are different
triangles that contain the silhouette of the object. Since the proxy fully contains the
original object, parts of the silhouette triangles may be transparent. This may cause
visible cracks when the object is rotated. Moving the clip planes reduces some of the
cracks.
If you move the plane that is behind the triangle farther away (using option –evlb),
some of the geometry that is behind the silhouette is included in the texture. When you
move the front clip plane closer to the cameras (using option –evlf), some of the
geometry that is in front of the silhouette is included in the texture.
Because some proxy triangles may have a texture with transparent edges, it may be
desirable to sort the proxy triangles. Because the proxy can be viewed from any direction,
007-1680-100
217
6: Creating Visual Effects
it is necessary to determine how the triangles are sorted. If the proxy is rendered with
only the nearest views selected, the triangles are ordered for each view differently. You
must set that mode using the option –nr. By default, three or four of the nearest views
are blended together. In that case, the proxy triangles are sorted for each group of views.
Sometimes it may be possible to see changes in transparency as the view moves from one
group of views to another. If this becomes too obvious, you can disable the sorting using
the option –evns.
Potential Problems
Generally, you can drastically reduce problems if you place the proxy very close to the
original object (especially around visible sharp edges) or if you increase the number of
views. The following are some potential problems you might encounter:
•
The images are missing an alpha channel.
If your machine does not support a single-buffered visual with at least 8 bits per
red, green, blue, and alpha component, the images may be missing an alpha
channel. Note the number of alpha bits printed at the beginning of the
makeProxyImages output. On some SGI systems with multisampling, you may
try to use the option –nms to request a visual without multisampling to improve the
probability of getting a visual with an alpha channel.
•
Your textures are not antialiased.
Do not forget to oversample the textures on machines with no antialiasing (using
option –o).
•
The processing time is very long.
The process may take a very long time if the proxy is very fine and many texture
pieces have to be added to each view. Since the rendering is done into a window,
ensure that you do not overlap the window during the process or that the screen
saver does not start. If some of the textures are corrupt, you may restart the program
with the same parameters and add the option –sfr, which skips the rendering of
the specified number of textures. It is also a good idea to increase shared arena size
(use environment variable PFSHAREDSIZE) to avoid memory overflow when the
pfIBRnode is saved at the end.
•
Texture pieces intersect the image of the object.
Inspecting the view textures, you may notice that sometimes the additional texture
pieces may intersect the image of the object. This is fine because those triangles that
are overlapped are assigned one of the additional texture pieces packed around the
object.
218
007-1680-100
Image-Based Rendering
Creating Images of an Object with makeIBRimages
You can use the program makeIBRimages from the directory
/usr/share/Performer/src/conv/ on IRIX and Linux and %PFROOT%\Src\conv
on Microsoft Windows to create images (views) of a specified object from a set of
directions. The input is a file that can be read by OpenGL Performer and the output is a
set of images of that object that can be directly used as an input for a pfIBRtexture. The
images are stored in a directory specified using the option -f.
If a text file info is present in the output directory, a set of 3D views is rendered. The file
has the same syntax as described in section “Creating a pfIBRtexture” on page 206. Each
line of the file info contains two values: the angle from the horizontal plane and how
many views are created for that angle. The images are then indexed by two integer values
that are appended to the name specified by the option -f. The first value is the ring index
of the views and the second one indexes the views within the ring.
If the file info is not present, a set of N views (set by the option -n) is computed around
the object using the horizontal angle of 0. In this case, only one index is appended to the
image name.
If you specify the option -pfb, the program outputs a pfb file in the specified directory.
The file contains a single pfIBRnode that uses the created images.
Note:
Before loading perfly, ensure that PFPATH is set to the directory that contains the
images.
If your machine does not support a single-buffered visual with at least 8 bits per red,
green, blue, and alpha component, the images may be missing the alpha channel. Note
the number of alpha bits printed when makeIBRimages begins.
When using pfIBRnodes and pfIBRtextures in perfly, you also need an alpha buffer. If
the pfIBRnode is rendered as a full rectangle, try the command-line parameter –9, in
which case perfly requests a visual with alpha.
To obtain the full set of command-line options, run the program makeIBRimages
without any parameters.
007-1680-100
219
6: Creating Visual Effects
Limitations
The following are current limitations of image-based rendering in OpenGL Performer:
220
•
A pfIBRtexture applied to a pfIBRnode is not properly rotated when the pfIBRnode
is viewed from the top. This may result in visible rotation of the texture with respect
to the ground.
•
When the flag PFIBR_3D_VIEWS is set in a pfIBRtexture, do not use 3D textures.
This mode is not implemented.
007-1680-100
Chapter 7
7. Importing Databases
Once you have learned how to create visual simulation applications with
OpenGL Performer, your next task is to import visual databases into those applications.
OpenGL Performer provides import and export functions for numerous popular
database formats to ease this effort.
This chapter describes the following:
•
The steps involved in creating custom loaders for other data formats
•
Pre-existing file loading utilities
•
Several utility functions in the OpenGL Performer database utility library that can
make the process of database conversion easier for you
•
Supported database formats
•
The Maya database exporter
Overview of OpenGL Performer Database Creation and Conversion
Source code is provided for most of the tools discussed in this chapter. In most cases the
loaders are short, easy to understand, and easy to modify.
Table 7-1 lists the subdirectories of /usr/share/Performer/src/lib on IRIX and
Linux and %PFROOT%\Src\lib on Microsoft Windows where you can find the source
code for the database processing tools.
Table 7-1
007-1680-100
Database-Importer Source Directories
Directory Name
Directory Contents
libpfdu
General database processing tools and utilities.
221
7: Importing Databases
Table 7-1
Database-Importer Source Directories (continued)
Directory Name
Directory Contents
libpfdb
Load, convert, and store specific database formats..
libpfutil
Additional utility functions.
Before you can import a database, you must create it. Some simulation applications
create data procedurally; for examples of this approach, see the “SGI PHD Format” on
page 268 or the “Sierpinski Sponge Loader” on page 280” sections of this chapter.
In most cases, however, you must create visual databases manually. Several software
packages are available to help with this task, and most such systems facilitate geometric
modeling, texture creation, and interactive specification of colors and material
properties. Some advanced systems support level-of-detail specification, animation
sequences, motion planning for jointed objects, automated roadway and terrain
generation, and other specialized functions.
- Utilities for Creating Efficient OpenGL Performer Run-Time
Structures
libpfdu
There are several layers of support in OpenGL Performer for loading 3D models and 3D
environments into OpenGL Performer run-time scene graphs. OpenGL Performer
contains the libpfdu library devoted to the import of data into (and export of data
from) OpenGL Performer run-time structures. Note that two database exporters have
already been written for the Medit and DWB database formats.
At the top level of the API, OpenGL Performer provides a standard set of functions to
read in files and convert databases of unknown type. This functionality is centered
around the notion of a database converter. A database converter is an abstract entity that
knows how to perform some or all of a set of database format conversion functions with
a particular database format. Moreover, converters must follow certain API guidelines
for standard functionality such that they can be easily integrated into OpenGL Performer
in a run-time environment without OpenGL Performer needing any prior knowledge of
a particular converter’s existence. This run-time integration is done through the use of
dynamic shared object (DSO) libraries on IRIX and Linux. On Microsoft Windows this is
accomplished using Dynamic-link Libraries (DLL).
222
007-1680-100
libpfdu - Utilities for Creating Efficient OpenGL Performer Run-Time Structures
pfdLoadFile - Loading Arbitrary Databases into OpenGL Performer
Table 7-2 describes the general routines for 3D databases provided by libpfdu.
Table 7-2
libpfdu Database Converter Functions
Function Name
Description
pfdInitConverter() Initialize the library and its classes for the desired format.
pfdLoadFile()
Load a database file into an OpenGL Performer scene graph.
pfdStoreFile()
Store a run-time scene graph into a database file.
pfdConvertFrom() Convert an external run-time format into an OpenGL Performer scene graph.
pfdConvertTo()
Convert an OpenGL Performer scene graph into an external run-time format.
The database loader utility library, libpfdu, provides a convenient function, named
pfdLoadFile(), that imports database files stored in any of the supported formats listed
in Table 7-6 on page 242.
Loading database files with pfdLoadFile() is easy. The function prototype is
pfNode *pfdLoadFile(char *fileName);
pfdLoadFile() tests the filename-extension portion of fileName (the substring starting at
the last period in fileName, if any) for one of the format-name codes listed in Table 7-6 on
page 242, then calls the appropriate importer.
The file-format selection process is implemented using dynamic loading of DSOs,
dynamic shared objects, for IRIX and Linux and DLLs, dynamic link libraries, for
Microsoft Windows. This process allows new loaders that are developed as database
formats change to be used with OpenGL Performer-based applications without
requiring recompilation of the OpenGL Performer application.
Note: Subsequent general references in this manual to DSOs also pertain to DLLs unless
otherwise noted.
If at all possible, pfdInitConverter() should be called before pfConfig() for the potential
formats that may be loaded. This will preload the DSO and allow it to initialize any of its
own data structures and classes. This is required if the loader DSO extends
007-1680-100
223
7: Importing Databases
OpenGL Performer classes or uses any node traversal callbacks so that if multiprocessing
these data elements will all have been precreated and be valid in all potential processes.
pfdInitConverter() automatically calls pfdLoadNeededDSOs_EXT() to preload
additional DSOs needed by the loader if the given loader has defined that routine. These
routines take a filename so that the loader has the option to search through the file for
possible DSO references in the file.
Loading Process Internals
The details of the loading process internal to pfdLoadFile() include the following:
1.
Searching for the named file using the current OpenGL Performer file path.
2. Extraction of the file-type extension.
3. Translation of the extension using a registered alias facility, formation of the DSO or
DLL name.
4. Formation of a loader function name.
5. Finding that function within the DSO using either dlsym() on IRIX and Linux or
GetProcAddress() on Microsoft Windows.
6. Searching first the current executable and loaded DSOs for the proper load function
and then searching through a list of user-defined and standard directories for that
DSO. Dynamic loading of the indicated DSO using dlopen() on IRIX and Linux and
using LoadLibrary() on Microsoft Windows.
7. Invocation of the loader function.
Loader Name
The loader function name is constructed from two components:
•
A prefix always consisting of pfdLoadFile_.
•
Loader suffix, which is the file extension string.
Note: The loader function pfdLoadFile_ must be exported using
_declspec(dllexport) on Microsoft Windows only.
224
007-1680-100
libpfdu - Utilities for Creating Efficient OpenGL Performer Run-Time Structures
Examples of several complete loader function names are shown in Table 7-3.
Table 7-3
Loader Name Composition
File Extension
Loader Function Name
dwb
pfdLoadFile_dwb()
flt
pfdLoadFile_flt()
medit
pfdLoadFile_medit()
obj
pfdLoadFile_obj()
pfb
pfdLoadFile_pfb()
Shell Environment Variables
Several shell environment variables are used in the loader location process. These are
PFLD_LIBRARY{N32,64}_PATH, LD_LIBRARY{N32,64}_PATH, and PFHOME.
Confusion about loader locations can be resolved by consulting the sources mentioned
earlier in this chapter to understand the use of these directory lists and reading the
following section, “Database Loading Details” on page 225. When the pfNotifyLevel is
set to the value for PFNFY_DEBUG (5) or greater, the DSO and loader function names
are printed as databases are loaded, as is the name of each directory that is searched for
the DSO.
The OpenGL Performer sample programs, including perfly, use pfdLoadFile() for
database importing. This allows them to simultaneously load and display databases in
many disparate formats. As you develop your own database loaders, follow the source
code examples in any of the libpfdb loaders. Then you will be able to load your data
into any OpenGL Performer application. You will not need to rebuild perfly or other
applications to view your databases.
Database Loading Details
Details about the database loading process are described further in this section, the
pfdLoadFile man page, and the source code which is in
/usr/share/Performer/src/lib/libpfdu/pfdLoadFile.c on IRIX and Linux
and in %PFROOT%\Src\lib\libpfdu\pfdLoadFile.c on Microsoft Windows.
007-1680-100
225
7: Importing Databases
The routines pfdInitConverter(), pfdLoadFile(), pfdStoreFile(), pfdConvertFrom(), and
pfdConvertTo() exist only as a level of indirection to allow you to manipulate all
databases regardless of format through a central API. They are in fact merely a
mechanism for creating an open environment for data sharing among the multitudes of
three-dimensional database formats. Each of these routines determines, using file-type
extensions, which database converter to load as a run-time DSO. The routine then calls
the appropriate functionality from that converter’s DSO. All converters must provide
API that is exactly the same as the corresponding libpfdu API with _EXT added to the
routine names (for example, for .medit files, the suffix is _medit). Note that multiple
physical extensions can be mapped to one converter extension with calls to
pfdAddExtAlias(). Several aliases are predefined upon initialization of libpfdu.
It is also important to note that because each of these converters is a unique entity that
they each may have state that is important to their proper function. Moreover, their
database formats may allow for multiple OpenGL Performer interpretations; so, there
exist APIs, shown in Table 7-4, not only to initialize and exit database converters, but also
to set and get modes, attributes, and values that might affect the converter’s
methodology.
Table 7-4
libpfdu Database Converter Management Functions
Function Name
Description
pfdInitConverter()
Initialize a database conversion DSO.
pfdExitConverter()
Exit a database conversion DSO.
pfdConverterMode()
Specify a mode for a specific conversion DSO.
pfdGetConverterMode()
Get a mode setting from a specific conversion DSO.
pfdConverterAttr()
Specify an attribute for a conversion DSO.
pfdGetConverterAttr()
Get an attribute setting from a conversion DSO.
pfdConverterVal()
Specify a value for a conversion DSO.
pfdGetConverterVal()
Get a value setting from a conversion DSO.
Once again each converter provides the equivalent routines with _EXT added to the
function name.
For example, the converter for the Open Inventor format would define the function
pfdInitConverter_iv() if it needed to be initialized before it was used. Likewise, it would
226
007-1680-100
libpfdu - Utilities for Creating Efficient OpenGL Performer Run-Time Structures
define the function pfdLoadFile_iv() to read an Open Inventor “.iv” file into an
OpenGL Performer scene graph.
Note: Because each converter is an individual entity (DSO) and deals with a particular
type of database, it may be the case that a converter will not provide all of the
functionality listed above, but rather only a subset. For instance, most converters that
come with OpenGL Performer only implement their version of pfdLoadFile but not
pfdStoreFile, pfdConvertFrom, or pfdConvertTo. However, users are free to add this
functionality to the converters using compliant APIs and OpenGL Performer’s libpfdu
will immediately recognize this functionality. Also, libpfdu traps access to nonexistent
converter functionality and returns gracefully to the calling code while notifying the user
that the functionality could not be found.
Finding and initializing a Converter
When one of the general database converter functions is called, it in turn calls the
corresponding routine provided by the converter, passing on the arguments it was given.
But the first time a converter is called, a search occurs to identify the converter and the
functions it provides. This is accomplished as follows.
•
Parse the extension—what appears after the final “.” in the filename. This is referred
to as EXT in the following bulleted items.
•
Check to see if any alias was created for the EXT extension with pfdAddExtAlias().
If a translation is defined, EXT is replaced with that extension.
•
Check the current executable to see if the symbol pfdLoadFile_EXT is already
defined, that is. if the loader was statically linked into the executable or a DSO was
previously loaded by some other mechanism. If not, the search continues.
For IRIX and Linux:
–
Generate a DSO library name to search for using the extension prototype
“libpfEXT_{-g,}.so”. This means the following strings will be constructed:
libpfEXT_.so for the optimized OpenGL loader
libpfEXT_-g.so for the debug OpenGL loader
–
Look for the DSO in several places, including the following:
.
$PFLD_LIBRARY_PATH
007-1680-100
227
7: Importing Databases
$LD_LIBRARY_PATH
$PFHOME/usr/lib{,32,64}/libpfdb
$PFHOME/usr/share/Performer/lib/libpfdb
–
Open the DSO using dlopen().
For Microsoft Windows:
–
Generate a DLL library name to search for using the extension prototype
“libpfEXT_{-g,}.so”. This means the following strings will be constructed:
libpfEXT_.so for the optimized OpenGL loader
libpfEXT_-g.so for the debug OpenGL loader
–
Look for the DLL in several places, including the following:
.
$PFLD_LIBRARY_PATH
$LD_LIBRARY_PATH
$PFHOME/Lib/libpfdb
$PFHOME/Lib/Debug/libpfdb
–
•
Open the DLL using LoadLibrary().
Once the object has been found, processing continues.
–
Query all libpfdu converter functionality from the symbol table of the DSO
using dlsym() on IRIX and Linux and of the DLL using GetProcAddress() on
Microsoft Windows with function names generated by appending _EXT to the
name of the corresponding pfd routine name. This symbol dictionary is
retained for future use.
–
Invoke the converter’s initialization function, pfdInitConverter_EXT(), if it
exists.
–
Invoke pfdLoadNeededDSOs_EXT() if it exists. This routine can then
recursively call pfdInitConverter_EXT(), as needed.
Developing Custom Importers
Having fully described how database converters can be integrated into OpenGL
Performer and the types of functionality they provide, the next undertaking is actually
implementing a converter from scratch. OpenGL Performer makes a great effort at
allowing the quick and easy development of effective and efficient database converters.
228
007-1680-100
Developing Custom Importers
While creating a new file loader for OpenGL Performer is not inherently difficult, it does
require a solid understanding of the following issues:
•
The structure and interpretation of the data file to be read
•
The scene graph concepts and nodes of libpf
•
The geometry and attribute definition objects of libpr
Structure and Interpretation of the Database File Format
In order to effectively convert a database into an OpenGL Performer scene graph, it is
important to have a substantial understanding of several concepts related to the original
database format:
•
The parsing of the file based on the database format
•
The data types represented in the format and their OpenGL Performer
correspondence
•
The scene graph structure of the file (if any)
•
The method of graphics state definition and inheritance defined in the format
Before trying to convert sophisticated 3D database formats into OpenGL Performer it is
important to have a thorough grasp of how every structure in the format needs to affect
how OpenGL Performer performs its run-time management of a scene graph. However,
although it requires a great deal of understanding to convert complex behaviors of
external formats into OpenGL Performer, it is still very straight forward to migrate basic
structure, geometry, and graphics state into efficient OpenGL Performer run-time
structures using the functionality provided in the OpenGL Performer database builder,
pfdBuilder.
Scene Graph Creation Using Nodes as Defined in libpf
Creating an OpenGL Performer scene graph requires a definite knowledge of the
following OpenGL Performer libpf node types: pfScene, pfGroup, and pfGeode.
These nodes can be used to define a minimally functional OpenGL Performer scene
graph. See “Nodes” in Chapter 3 for more details on libpf and OpenGL Performer
scene graphs and node types.
007-1680-100
229
7: Importing Databases
Defining Geometry and Graphics State for libpr
In order to input geometry and graphics into OpenGL Performer, it is important to have
an understanding of how OpenGL Performer’s low-level rendering objects work in
libpr, OpenGL Performer’s performance rendering library. The main libpr
rendering primitives are a pfGeoSet and a pfGeoState. A pfGeoSet is a collection of like
geometric primitives that can all be rendered in exactly the same way in one large
continuous chunk. A pfGeoState is a complete definition of graphics mode settings for
the rendering hardware and software. It contains many attributes such as texture and
material. Given a pfGeoSet and a corresponding pfGeoState, libpr can completely and
efficiently render all of the geometry in the pfGeoSet. For a more detailed description of
pfGeoSets and pfGeoStates, see “pfGeoSets and pfGeoStates” in Chapter 12, which goes
into detail on all libpr primitives and how OpenGL Performer will use them.
However, realizing that OpenGL Performer’s structuring of geometry and graphics state
is optimized for rendering speed and not for modeling ease or general conceptual
partitioning, OpenGL Performer now contains a new mechanism for translating external
graphics state and geometry into efficient libpr structures. This new mechanism is the
pfdBuilder that exists in libpfdu.
The pfdBuilder allows the immediate mode input of graphics state and primitives
through very simple and exposed data structures. After having received all of the
relevant information, the pfdBuilder builds efficient and somewhat optimized libpr
data structures and returns a low-level libpf node that can be attached to an OpenGL
Performer scene graph. The pfdBuilder is the recommended method of importing data
from non-OpenGL Performer-based formats into OpenGL Performer.
Creating an OpenGL Performer Database Converter using libpfdu
Creating a new format converter is very simple process. More than thirty database
loaders are shipped with OpenGL Performer in source code form to serve as practical
examples of this process. The loaders read formats that range from trivial to complex and
should serve as an instructive starting point for those developing loaders for other
formats. These loaders can be found in the directory
/usr/share/Performer/src/lib/libpfdb/libpf* on IRIX and Linux and in
%PFROOT%\Src\lib\libpfdb\libpf* on Microsoft Windows.
This section describes the libpfdu framework for creating a 3D database format
converter. Consider writing a converter for a simple ASCII format that is called the
Imaginary Immediate Mode format with the file type extension .iim. This format is
230
007-1680-100
Developing Custom Importers
much like the more elaborate .im format loader used at SGI for the purposes of testing
basic OpenGL Performer functionality.
The first thing to do is set up the routine that pfdLoadFile() will call when it attempts to
load a file with the extension .iim.
#ifdef WIN32
#define PFDB_DLLEXPORT __declspec(dllexport)
#else
#define PFDB_DLLEXPORT /* no-op */
#endif
extern PFDB_DLLEXPORT pfNode *pfdLoadFile_iim(char *fileName)
{
}
This function needs to perform several basic actions:
1.
Find and open the given file.
2. Reset the libpfdu pfdBuilder for input of new geometry and state.
3. Set up any pfdBuilder modes that the converter needs enabled.
4. Set up local data structures that can be used to communicate geometry and graphics
state with the pfdBuilder.
5. Set up a libpf pfGroup which can hold all of the logical partitions of geometry in
the file (or hold a subordinate collection of nodes as a general scene graph if the
format supports it).
6. Optionally set up a default state to use for geometry with unspecified graphics state.
7. Parse the file, which entails the following:
007-1680-100
•
Filling in the local geometry and graphics state data structures
•
Passing them to the pfdBuilder as inputted from the file
•
Asking the pfdBuilder to build the data structures into OpenGL Performer data
structures when a logical partition of the file has ended
•
Attaching the OpenGL Performer node returned by the build to the higher-level
group which will hold the entire OpenGL Performer representation of this file.
Note that this step becomes more complex if the format supports the notion of
hierarchy only in that the appropriate libpf nodes must be created and
attached to each other using pfAddChild() to build the hierarchy. In this case
231
7: Importing Databases
requests are made for the builder to build after inputting all of the geometry
and state found in a particular leaf node in the database.
8. Delete local data structures used to input geometry and graphics state.
9. Close the file.
10. Perform any optional optimization of the OpenGL Performer scene graph.
Optimizations might include calls to pfdFreezeTransforms(), pfFlatten() or
pfdCleanTree().
11. Return the pfGroup containing the entire OpenGL Performer representation of the
database file.
Steps 1-8 expand the function outline to the following:
extern PFDB_DLLEXPORT pfNode *pfdLoadFile_iim(char *fileName)
{
FILE* iimFile;
pfdGeom* polygon;
pfGroup* root;
/* Performer has utility for finding and opening file */
if ((iimFile = pfdOpenFile(fileName)) == NULL)
return NULL;
/* Clear builder from previous converter invocations */
pfdResetBldrGeometry();
pfdResetBldrState();
/* Call pfdBldrMode for any needed modes here */
/* Create polygon structure */
/* holds one N-sided polygon where N is < 300 */
polygon = pfdNewGeom(300);
/* Create pfGroup to hold entire database */
/* loaded from this file */
root = pfNewGroup();
/* Specify state for geometry with no graphics state */
/* As well as default enables, etc. This routine */
/* should invoke pfdCaptureDefaultBldrState()*/
SetupDefaultGraphicsStateIfThereIsOne();
/* Do all the real work in parsing the file and */
232
007-1680-100
Developing Custom Importers
/* converting into Performer */
ParseIIMFile(iimFile, root, polygon);
/* Delete local polygon struct */
pfdDelGeom(polygon);
/* Close File */
fclose(iimFile);
/* Optimize OpenGL Performer scene graph */
/* via use of pfFlatten, pfdCleanTree, etc. */
OptimizeGraph(root);
return (pfNode*)root;
}
At the heart of the file loader lies the ParseIIMFile() function. The specifics of parsing a
file are completely dependent on the format; so, the parsing will be left as an exercise to
you. However, the following code fragments should show a framework for what goes
into integrating the parser with the pfdBuilder framework for geometry and graphics
state data conversion. Note that several possible graphics state inheritance models might
be used in external formats and that the pfdBuilder is designed to support all of them:
•
The default pfdBuilder state inheritance is that of immediate mode graphics state.
Immediate mode state is specified through calls to pfdBldrStateMode(),
pfdBldrStateAttr(), and pfdBldrStateVal().
•
There also exists a pfdBuilder state stack for hierarchical state application to
geometry. This is accomplished through the use of pfdPushBldrState() and
pfdPopBldrState() in conjunction with the normal use of the immediate mode
pfdBuilder state API.
•
Lastly, there is a pfdBuilder named state list that can be used to define a number of
"named materials" or "named state definitions" that can then be recalled in one API
called (for instance, you might define a "brick" state with a red material and a brick
texture. Later you might just want to say "brick" is the current state and then input
the walls of several buildings). This type of state naming is accomplished by fully
specifying the state to be named using the immediate mode API and then calling
pfdSaveBldrState(). This state can then be recalled using pfdLoadBldrState().
ParseIIMFile(FILE *iimFile, pfGroup *root, pfdGeom *poly)
{
while((op = GetNextOp(iimFile)) != NULL)
{
switch(op)
007-1680-100
233
7: Importing Databases
{
case GEOMETRY_POLYGON:
polygon->numVerts = GetNumVerts(iimFile);
/* Determine if polygon has Texture Coords */
if (pfdGetBldrStateMode(PFSTATE_ENTEXTURE)==PF_ON)
polygon->tbind = PFGS_PER_VERTEX;
else
polygon->tbind = PFGS_OFF;
/* Determine if Polygon has normals */
if (AreThereNormalsPerVertex() == TRUE)
polygon->nbind = PFGS_PER_VERTEX;
else if
(pfdGetBldrStateMode(PFSTATE_ENLIGHTING)==PF_ON)
polygon->nbind = PFGS_PER_PRIM;
else
polygon->nbind = PFGS_OFF;
/* Determine if Polygon has colors */
if (AreThereColorsPerVertex() == TRUE)
polygon->cbind = PFGS_PER_VERTEX;
else if (AreThereColorsPerPrim() == TRUE)
polygon->cbind = PFGS_PER_PRIM;
else
polygon->cbind = PFGS_OFF;
for(i=0;i<polygon->numVerts;i++)
{
/* Read ith Vertex into local data structure */
polygon->coords[i][0] = GetNextVertexFloat();
polygon->coords[i][1] = GetNextVertexFloat();
polygon->coords[i][2] = GetNextVertexFloat();
/* Read texture coord for ith vertex if any */
if (polygon->tbind == PFGS_PER_VERTEX)
{
polygon->texCoords[i][0] = GetNextTexFloat();
polygon->texCoords[i][1] = GetNextTexFloat();
}
/* Read normal for ith Vertex if normals bound*/
if (polygon->nbind == PFGS_PER_VERTEX)
{
polygon->norms[i][0] = GetNextNormFloat();
polygon->norms[i][1] = GetNextNormFloat();
234
007-1680-100
Developing Custom Importers
polygon->norms[i][2] = GetNextNormFloat();
}
/* Read only one normal per prim if necessary */
else if ((polygon->nbind == PFGS_PER_PRIM) &&
(i == 0))
{
polygon->norms[0][0] = GetNextNormFloat();
polygon->norms[0][1] = GetNextNormFloat();
polygon->norms[0][2] = GetNextNormFloat();
}
/* Get Color for the ith Vertex if color bound*/
if (polygon->cbind == PFGS_PER_VERTEX)
{
polygon->colors[i][0] =
GetNextColorFloat();
polygon->colors[i][1] =
GetNextColorFloat();
polygon->colors[i][2] =
GetNextColorFloat();
}
/* Get one color per prim if necessary */
else if ((polygon->cbind == PFGS_PER_PRIM) &&
(i == 0))
{
polygon->colors[0][0] =
GetNextColorFloat();
polygon->colors[0][1] =
GetNextColorFloat();
polygon->colors[0][2] =
GetNextColorFloat();
}
}
/* Add this polygon to pfdBuilder */
/* Because it is a single poly, 1 */
/* is specified here */
pfdAddBldrGeom(1);
break;
case GRAPHICS_STATE_TEXTURE:
{
char *texName;
pfTexture *tex;
texName = ReadTextureName(iimFile);
if (texName != NULL)
{
007-1680-100
235
7: Importing Databases
/* Get prototype tex from pfdBuilder*/
tex =
pfdGetTemplateObject(pfGetTexClassType());
/* This clears that object to default */
pfdResetObject(tex);
/* If just the name of a pfTexture is */
/* set, pfdBuilder will auto find & Load */
/* the texture*/
pfTexName(tex,texName);
/* This is the current pfdBuilder */
/* texture and texturing is on */
pfdBldrStateAttr(PFSTATE_TEXTURE,tex);
pfdBldrStateMode(PFSTATE_ENTEXTURE, PF_ON);
}
else
{
/* No texture means disable texturing */
/* And set current texture to NULL */
pfdBldrStateMode(PFSTATE_ENTEXTURE,PF_OFF);
pfdBldrStateAttr(PFSTATE_TEXTURE, NULL);
}
}
break;
case GRAPHICS_STATE_MATERIAL:
{
pfMaterial *mtl;
mtl = pfdGetTemplateObject(pfGetMtlClassType());
pfdResetObject(mtl);
pfMtlColor(mtl, PFMTL_AMBIENT,
GetAmRed(), GetAmGreen(), GetAmBlue());
pfMtlColor(mtl, PFMTL_DIFFUSE,
GetDfRed(), GetDfGreen(), GetDfBlue());
pfMtlColor(mtl, PFMTL_SPECULAR,
GetSpRed(), GetSpGreen(), GetSpBlue());
pfMtlShininess(mtl, GetMtlShininess());
pfMtlAlpha(mtl, GetMtlAlpha());
pfdBldrStateAttr(PFSTATE_FRONTMTL, mtl);
pfdBldrStateAttr(PFSTATE_BACKMTL, mtl);
}
break;
case GRAPHICS_STATE_STORE:
pfdSaveBldrState(GetStateName());
236
007-1680-100
Developing Custom Importers
break;
case GRAPHICS_STATE_LOAD:
pfdLoadBldrState(GetStateName());
break;
case GRAPHICS_STATE_PUSH:
pfdPushBldrState();
break;
case GRAPHICS_STATE_POP:
pfdPopBldrState();
break;
case GRAPHICS_STATE_RESET:
pfdResetBldrState();
break;
case GRAPHICS_STATE_CAPTURE_DEFAULT:
pfdCaptureDefaultBldrState();
break;
case BEGIN_LEAF_NODE:
/* Not really necessary because it is */
/* destroyed on build*/
pfdResetBldrGeometry();
break;
case END_LEAF_NODE:
{
pfNode *nd = pfdBuild();
if (nd != NULL)
pfAddChild(root,nd);
}
break;
}
}
}
One of the fundamental structures involved in the above routine outline is the pfdGeom
structure which you fill in with information about a single primitive, or a single strip of
primitives. The pfdGeom structure is essential in communicating with the pfdBuilder
and is defined as follows:
typedef struct _pfdGeom
{
int
flags;
int
nbind, cbind, tbind[PF_MAX_TEXTURES];
int
short
float
007-1680-100
numVerts;
primtype;
pixelsize;
237
7: Importing Databases
/* Non-indexed attributes - do not set if poly is indexed */
pfVec3
*coords;
pfVec3
*norms;
pfVec4
*colors;
pfVec2
*texCoords[PF_MAX_TEXTURES];
/* Indexed attributes - do not set if poly is non-indexed */
pfVec3
*coordList;
pfVec3
*normList;
pfVec4
*colorList;
pfVec2
*texCoordList[PF_MAX_TEXTURES];
/* Index lists - do not set if poly is non-indexed */
ushort
*icoords;
ushort
*inorms;
ushort
*icolors;
ushort
*itexCoords[PF_MAX_TEXTURES];
int
numTextures;
struct _pfdGeom
*next;
} pfdGeom;
See the pfdGeoBuilder(3pf) man pages for more information on using this structure
along with its sister structure, the pfdPrim.
The above should provide a well-defined framework for creating a database converter
that can be used with any OpenGL Performer applications using the pfdLoadFile()
functionality.
However, it is also important to note that there are a multitude of pfdBuilder modes and
attributes that can be used to affect some of the basic methods that the builder actually
uses:
238
007-1680-100
Developing Custom Importers
Table 7-5
pfdBuilder Modes and Attributes
Function Name
Token Description
pfd{Get}BldrMode()
PFDBLDR_MESH_ENABLE
PFDBLDR_MESH_SHOW_TSTRIPS
PFDBLDR_MESH_INDEXED
PFDBLDR_MESH_MAX_TRIS
PFDBLDR_MESH_RETESSELLATE
PFDBLDR_MESH_LOCAL_LIGHTING
PFDBLDR_AUTO_COLORS
PFDBLDR_AUTO_NORMALS
PFDBLDR_AUTO_ORIENT
PFDBLDR_AUTO_ENABLES
PFDBLDR_AUTO_CMODE
PFDBLDR_AUTO_DISABLE_TCOORDS_BY_STATE
PFDBLDR_AUTO_DISABLE_NCOORDS_BY_STATE
PFDBLDR_AUTO_LIGHTING_STATE_BY_NCOORDS
PFDBLDR_AUTO_LIGHTING_STATE_BY_MATERIALS
PFDBLDR_AUTO_TEXTURE_STATE_BY_TEXTURES
PFDBLDR_AUTO_TEXTURE_STATE_BY_TCOORDS
PFDBLDR_BREAKUP
PFDBLDR_BREAKUP_SIZE
PFDBLDR_BREAKUP_BRANCH
PFDBLDR_BREAKUP_STRIP_LENGTH
PFDBLDR_SHARE_MASK
PFDBLDR_ATTACH_NODE_NAMES
007-1680-100
239
7: Importing Databases
Table 7-5
pfdBuilder Modes and Attributes (continued)
Function Name
Token Description
PFDBLDR_DESTROY_DATA_UPON_BUILD
PFDBLDR_PF12_STATE_COMPATIBLE
PFDBLDR_BUILD_LIMIT
PFDBLDR_GEN_OPENGL_CLAMPED_TEXTURE_COORDS
PFDBLDR_OPTIMIZE_COUNTS_NULL_ATTRS
pfd{Get}BldrAttr()
PFDBLDR_NODE_NAME_COMPARE
PFDBLDR_STATE_NAME_COMPARE
Because the pfdBuilder is released as source code, it is easy to add further functionality
and more modes and attributes to even further customize this central functionality.
In fact, because the pfdBuilder acts as a “data funnel” in converting data into OpenGL
Performer run-time structures, it is easy to control the behavior of many standard
conversion tasks through merely globally setting builder modes which will subsequently
affect all converters that use the pfdBuilder to process their data.
Maximizing Database Loading and Paging Performance with PFB and PFI
Formats
“Description of Supported Formats” on page 244 describes all of the file formats
supported by OpenGL Performer. Although you can use files in these formats directly,
you can dramatically reduce database loading time by preconverting databases into the
PFB format and images into the PFI format.
To convert to the PFB file format or the PFI image format, use the pfconv and pficonv
utilities.
pfconv
The pfconv utility converts from any format for which a pfdLoadFile...() function exists
into any format for which a pfdStoreFile...() exists. The most common format to convert
240
007-1680-100
Supported Database Formats
to is the PFB format. For example, to convert cow.obj into the PFB format, use the
following command:
% pfconv cow.obj cow.pfb
By default, pfconv optimizes the scene graph when doing the conversion. The
optimizations are controlled with the -o and -O command line options. Builder options
are controlled with the -b and -B command line options. Converter modes are
controlled with the -m and -M command line options. Refer to the help page for more
specific information about the command line options by entering:
% pfconv -h
Example Conversion
When converting to the PFB format, texture files can be converted to the PFI format using
the following command line options:
% pfconv -M pfb, 5, 1
5 means PFPFB_SAVE_TEXTURE_PFI.
1 means convert .rgb texture images to .pfi.
pficonv
The pficonv utility converts from IRIS libimage format to PFI format image files. For
example, to convert cafe.rgb into the PFI format, use the following command:
% pficonv cafe.rgb cafe.pfi
MIPmaps can be automatically generated and stored in the resulting PFI files by adding
-m to the command line.
Supported Database Formats
Vendors of several leading database construction and processing tools have provided
database-loading software for you to use with OpenGL Performer. This section describes
these loaders, the loaders developed by the OpenGL Performer engineering team and
007-1680-100
241
7: Importing Databases
several loaders developed in the OpenGL Performer user community for other database
formats.
Importing your databases is simple if they are in formats for which OpenGL Performer
database loaders have already been written. Each of the loaders listed in Table 7-6 is
included with OpenGL Performer. If you want to import or export databases in any of
these formats, refer to the appropriate section of this chapter for specific details about the
individual loaders.
242
Table 7-6
Supported Database Formats
Name
Description
3ds
AutoDesk 3DStudio binary data
bin
SGI format used by powerflip
bpoly
Side Effects Software PRISMS binary data
byu
Brigham Young University CAD/FEA data
csb
OpenGL Optimizer Format
ct
Cliptexture config file loader - auto-generates viewing geometry
dwb
Coryphaeus Software Designer’s Workbench data
dxf
AutoDesk AutoCAD ASCII format
flt11
MultiGen public domain Flight v11 format
flt
MultiGen OpenFlight format provided by MultiGen
gds
McDonnell-Douglas GDS things data
gfo
Old SGI radiosity data format
im
Simple OpenGL Performer data format
irtp
AAI/Graphicon Interactive Real-Time PHIGS
iv
SGI Open Inventor format (VRML 1.0 superset)
lsa
Lightscape Technologies ASCII radiosity data
lsb
Lightscape Technologies binary radiosity data
007-1680-100
Supported Database Formats
007-1680-100
Table 7-6
Supported Database Formats (continued)
Name
Description
medit
Medit Productions medit modeling data
nff
Eric Haines’ ray tracing test data
pfb
OpenGL Performer fast binary format
obj
Wavefront Technologies data format
pegg
Radiosity research data format
phd
SGI polyhedron data format
poly
Side Effects Software PRISMS ASCII data
ptu
Simple OpenGL Performer terrain data format
rpc
ArchVision rich photorealistic content
sgf
US Naval Academy standard graphics format
sgo
Paul Haeberli’s graphics data format
spf
US Naval Academy simple polygon format
sponge
Sierpinski sponge 3D fractal generator
star
Astronomical data from Yale University star chart
stla
3D Structures ASCII stereolithography data
stlb
3D Structures binary stereolithography data
stm
Michael Garland’s terrain data format
sv
John Kichury’s i3dm modeler format
tri
University of Minnesota Geometry Center data
unc
University of North Carolina walkthrough data
wrl
OpenWorlds VMRL 2.0 provided by DRaW Computing
243
7: Importing Databases
Description of Supported Formats
This section describes the different database file formats that OpenGL Performer
supports.
AutoDesk 3DS Format
The AutoDesk 3DS format is used by the 3DStudio program and by a number of 3D
file-interchange tools. The OpenGL Performer loader for 3DS files is located in the
directory /usr/share/Performer/src/lib/libpfdb/libpf3ds on IRIX and
Linux and in %PFROOT%\Src\lib\libpfdb\libpf3ds on Microsoft Windows. This
loader uses an auxiliary library, 3dsftk.a, to parse and interpret the 3ds file.
pfdLoadFile() uses the function pfdLoadFile_3ds() to import data from 3DStudio files
into OpenGL Performer run-time data structures.
SGI BIN Format
The SGI BIN format is supported by both Showcase and the powerflip demonstration
program. BIN files are in a simple format that specifies only independent quadrilaterals.
The image in Figure 7-1 shows several of the BIN-format objects provided in the OpenGL
Performer sample data directory.
244
007-1680-100
Description of Supported Formats
Figure 7-1
BIN-Format Data Objects
The source code for the BIN-format importer pfdLoadFile_bin() is provided in the file
pfbin.c. This code shows how easy it can be to implement an importer. Since
pfdLoadFile_bin() is based on the pfdBuilder() utility function, it will build efficient
triangle-strip pfGeoSets from the quadrilaterals of a given BIN file. The BIN format has
the following structure:
1.
A 4-byte magic number, 0x5432, which identifies the file as a BIN file.
2. A 4-byte number that contains the number of vertices, which is four times the
number of quadrilaterals.
3. Four bytes of zero.
4. A list of polygon data for each vertex in the object. The data consists of three
floating-point words of information about normals followed by three floating-point
words of vertex information.
The BIN format uses these data structures:
typedef struct
{
float normal[3];
float coordinate[3];
007-1680-100
245
7: Importing Databases
} Vertex;
typedef struct
{
long magic;
long vertices;
long zero;
Vertex vertex[1];
} BinFile;
pfdLoadFile() uses the function pfdLoadFile_bin() to import data from BIN format files
into OpenGL Performer run-time data structures:
The pfdLoadFile_bin() function composes a random color for each file it reads. The
chosen color has red, green, and blue components uniformly distributed within the range
0.2 to 0.7 and is fully opaque.
Side Effects POLY Format
The Side Effects software PRISMS database modeler format supports both ASCII and
binary forms of the POLY format. The OpenGL Performer loader for ASCII “.poly” files
is located in the directory
/usr/share/Performer/src/lib/libpfdb/libpfpoly for IRIX and Linux and
in %PFROOT%\Src\lib\libpfdb\libpfpoly for Microsoft Windows. The binary
format “.bpoly” loader is located in the directory
/usr/share/Performer/src/lib/libpfdb/libpfbpoly for IRIX and Linux and
in %PFROOT%\Src\lib\libpfdb\libpfbpoly for Microsoft Windows. These
formats are equivalent in content and differ only in representation.
The POLY format is an easy to understand ASCII data representation with the following
structure:
1.
A text line containing the keyword “POINTS”
2. One text line for each vertex in the file. Each line begins with a vertex number,
followed by a colon, followed by the X, Y, and Z axis coordinates of the vertex,
optional additional information, and a new-line character. The optional information
includes color specification in the form “c(R,G,B,A)”, a normal vector of the form
“n(NX,NY,NZ)”, or a texture coordinate in the form “uv(S,T)” where each of the
values shown are floating point numbers.
3. A text line containing the keyword “POLYS”
246
007-1680-100
Description of Supported Formats
4. One text line for each polygon in the file. Each line begins with a polygon number,
followed by a colon, followed by a series of vertex indices, optional additional
information, an optional “<“character, and a new-line. The optional information
includes color specification in the form “c(R,G,B,A)”, a normal vector of the form
“n(NX,NY,NZ)”, or a texture coordinate in the form “uv(S,T)” where the values in
parentheses are floating point numbers.
Here is a sample POLY format file for a cube with colors, texture coordinates, and
normals specified at each vertex:
POINTS
1: -0.5 -0.5 -0.5 c(0, 0, 0, 1) uv(0, 0) n(0, -1, 0)
2: -0.5 -0.5 0.5 c(0, 0, 1, 1) uv(0, 0) n(0, -1, 0)
3: 0.5 -0.5 0.5 c(1, 0, 1, 1) uv(1, 0) n(0, -1, 0)
4: 0.5 -0.5 -0.5 c(1, 0, 0, 1) uv(1, 0) n(0, -1, 0)
5: -0.5 -0.5 0.5 c(0, 0, 1, 1) uv(0, 0) n(0, 0, 1)
6: -0.5 0.5 0.5 c(0, 1, 1, 1) uv(0, 1) n(0, 0, 1)
7: 0.5 0.5 0.5 c(1, 1, 1, 1) uv(1, 1) n(0, 0, 1)
8: 0.5 -0.5 0.5 c(1, 0, 1, 1) uv(1, 0) n(0, 0, 1)
9: -0.5 0.5 0.5 c(0, 1, 1, 1) uv(0, 1) n(0, 1, 0)
10: -0.5 0.5 -0.5 c(0, 1, 0, 1) uv(0, 1) n(0, 1, 0)
11: 0.5 0.5 -0.5 c(1, 1, 0, 1) uv(1, 1) n(0, 1, 0)
12: 0.5 0.5 0.5 c(1, 1, 1, 1) uv(1, 1) n(0, 1, 0)
13: -0.5 -0.5 -0.5 c(0, 0, 0, 1) uv(0, 0) n(0, 0, -1)
14: 0.5 -0.5 -0.5 c(1, 0, 0, 1) uv(1, 0) n(0, 0, -1)
15: 0.5 0.5 -0.5 c(1, 1, 0, 1) uv(1, 1) n(0, 0, -1)
16: -0.5 0.5 -0.5 c(0, 1, 0, 1) uv(0, 1) n(0, 0, -1)
17: -0.5 -0.5 -0.5 c(0, 0, 0, 1) uv(0, 0) n(-1, 0, 0)
18: -0.5 0.5 -0.5 c(0, 1, 0, 1) uv(0, 1) n(-1, 0, 0)
19: -0.5 0.5 0.5 c(0, 1, 1, 1) uv(0, 1) n(-1, 0, 0)
20: -0.5 -0.5 0.5 c(0, 0, 1, 1) uv(0, 0) n(-1, 0, 0)
21: 0.5 0.5 0.5 c(1, 1, 1, 1) uv(1, 1) n(1, 0, 0)
22: 0.5 0.5 -0.5 c(1, 1, 0, 1) uv(1, 1) n(1, 0, 0)
23: 0.5 -0.5 -0.5 c(1, 0, 0, 1) uv(1, 0) n(1, 0, 0)
24: 0.5 -0.5 0.5 c(1, 0, 1, 1) uv(1, 0) n(1, 0, 0)
POLYS
1: 1 2 3 4 <
2: 5 6 7 8 <
3: 9 10 11 12 <
4: 13 14 15 16 <
5: 17 18 19 20 <
6: 21 22 23 24 <
007-1680-100
247
7: Importing Databases
pfdLoadFile() uses the functions pfdLoadFile_poly() and pfdLoadFile_bpoly() to
import data from “.poly” and “.bpoly” format files into OpenGL Performer run-time
data structures.
Brigham Young University BYU Format
The Brigham Young University “.byu” format is used as an interchange format by some
finite element analysis packages. The OpenGL Performer loader for “.byu” files is
located in the directory /usr/share/Performer/src/lib/libpfdb/libpfbyu for
IRIX and Linux and in %PFROOT%\Src\lib\libpfdb\libpfbyu for Microsoft
Windows.
The format of a BYU file consists of four parts as defined below:
1.
A text line containing four counts: the number of parts, the number of vertices,
the number of polygons, and the number of elements in the connectivity array.
2. The part definition list, containing the starting polygon number and ending
polygon number (one pair per line) for parts lines.
3. The vertex list, which has the X, Y, Z coordinates of each vertex in the database
packed two per line. This means that vertices 1 and 2 are on the first line, 3 and 4 are
on the second, and so on for (vertices + 1)/2 lines of text in the file.
4. The connectivity array, with an entry for each polygon. These entries may span
multiple lines in the input file and each consists of three or more vertex indices with
the last negated as an end of list flag. For example, if the first polygon were a quad,
the connectivity array might start with “1 2 3 -4” to define a polygon that connects
the first four vertices in order.
The following BYU format file defines two adjoining quads:
2 6 2
1 1
2 2
0 0 0
10 10
10 10
1 2 3
4 3 5
0
10 0 0
0 0 10 0
10 0 10 10
-4
-6
pfdLoadFile() uses the function pfdLoadFile_byu() to import data from “.byu” format
files into OpenGL Performer run-time data structures.
248
007-1680-100
Description of Supported Formats
Optimizer CSB Format
OpenGL Performer can load native OpenGL Optimizer format files using this loader.
OpenGL Optimizer can also load OpenGL Performer’s PFB native format files,
providing full database interoperability. This allows you to use OpenGL Optimizer
database simplification and optimization tools on OpenGL Performer databases.
Virtual Cliptexture CT Loader
The OpenGL Performer CT loader allows you to create and configure cliptextures and
virtual cliptextures, complete with a scene graph containing simple geometry and
callbacks. See the Cliptexture chapter for more details.
Designer’s Workbench DWB Format
The binary DWB format is used for input and output by the Designer’s Workbench,
EasyT, and EasyScene database modeling tools produced by Coryphaeus Software. DWB
is an advanced database format that directly represents many of OpenGL Performer’s
attribute and hierarchical scene graph concepts.
An importer for this format, named pfdLoadFile_dwb(), has been provided by
Coryphaeus Software for your use. The loader code and its associated documentation are
in the directory /usr/share/Performer/src/lib/libpfdb/libpfdwb for IRIX
and Linux and in %PFROOT%\Src\lib\libpfdb\libpfdwb for Microsoft
Windows.The image in Figure 7-2 shows a model of the Soma Cube puzzle invented by
Piet Hein. The model was created using Designer’s Workbench. Each of the pieces is
stored as an individual DWB-format file. Do you see how to form the 3 x 3 cube at the
lower left from the seven individual pieces?
007-1680-100
249
7: Importing Databases
Figure 7-2
Soma Cube Puzzle in DWB Form
pfdLoadFile() uses the function pfdLoadFile_dwb() to load Designer’s Workbench files
into OpenGL Performer run-time data structures.
AutoCAD DXF Format
The DXF format originated with Autodesk’s AutoCAD database modeling system. The
version recognized by the pfdLoadFile_dxf() database importer is a subset of ASCII
Drawing Interchange Format (DXF) Release 12. The binary version of the DXF format,
also known as DXF, is not supported. Source code for the importer is in the file
/usr/share/Performer/src/lib/libpfdb/libpfdxf/pfdxf.c for IRIX and
Linux and in %PFROOT%\Src\lib\libpfdb\libpfdxf\pfdxf.c for Microsoft
Windows. pfdLoadFile_dxf() was derived from the DXF-to-DKB data file converter
developed and placed in the public domain by Aaron A. Collins.
The image in Figure 7-3 shows a DXF model of the famous Utah teapot. This model was
loaded from DXF format using the pfdLoadFile_dxf() database importer.
250
007-1680-100
Description of Supported Formats
Figure 7-3
The Famous Teapot in DXF Form
The DXF format has an unusual though well-documented structure. The general
organization of a DXF file is the following:
1.
HEADER section with general information about the file
2. TABLES section to provide definitions for named items, including:
007-1680-100
■
LTYPE, the line-type table
■
LAYER, the layer table
■
STYLE, the text-style table
■
VIEW, the view table
■
UCS, the user coordinate-system table
■
VPORT, the viewport configuration table
■
DIMSTYLE, the dimension style table
■
APPID, the application identification table
251
7: Importing Databases
3. BLOCKS section containing block definition entities
4. ENTITIES section containing entities and block references
5. END-OF-FILE
Within each section are groups of values, where each value is defined by a two-line pair
of tokens. The first token is a numeric code indicating how to interpret the information
on the next line. For example, the sequence
10
1.000
20
5.000
30
3.000
defines a “start point” at the XYZ location (1, 5, 3). The codes 10, 20, and 30 indicate,
respectively, that the primary X, Y, and Z values follow. All data values are retained in a
set of numbered registers (10, 20, and 30 in this example), which allows values to be
reused. This simple state-machine type of run-length coding makes DXF files
space-efficient at the cost of making them harder to interpret.
pfdLoadFile() uses the function pfdLoadFile_dxf() to load DXF format files into
OpenGL Performer run-time data structures.
Several widely available technical books provide full details of this format if you need
more information. Chief among these are AutoCAD Programming, 2nd Edition, by Dennis
N. Jump, Windcrest Books, 1991, and AutoCAD: The Complete Reference, Second Edition, by
Nelson Johnson, Osborne McGraw-Hill, 1991.
MultiGen OpenFlight Format
The OpenFlight format is a binary format used for input and output by the MultiGen and
ModelGen database modeling tools produced by MultiGen. It is a comprehensive format
that can represent nearly all of OpenGL Performer’s advanced concepts, including object
hierarchy, instancing, level-of-detail selection, light-point specification, texture mapping,
and material property specification.
MultiGen has provided an OpenFlight-format importer, pfdLoadFile_flt(), for your use.
The loaders and associated documentation are in the directories
/usr/share/Performer/src/lib/libpfdb/libpfflt11 and libpfflt for
252
007-1680-100
Description of Supported Formats
IRIX and Linux and in %PFROOT%\Src\lib\libpfdb\libpfflt11 and libpfllt
for Microsoft Windows. Refer to the Readme files in these directories for important
information about the loaders and for help in contacting MultiGen for information about
pfdLoadFile_flt() or the OpenFlight format.
The image in Figure 7-4 shows a model of a spacecraft created by Viewpoint Animation
Engineering using MultiGen. This OpenFlight format model was loaded into OpenGL
Performer using pfdLoadFile_flt().
Figure 7-4
Spacecraft Model in OpenFlight Format
pfdLoadFile() uses the function pfdLoadFile_flt() to load OpenFlight format files into
OpenGL Performer run-time data structures.
Files in the OpenFlight format are structured as a linear sequence of records. The first few
bytes of each record are a header containing an op-code, the length of the record, and
possibly an ASCII name for the record. The first record in the file is a special “database
header” record whose op-code, stored as a 2-byte short integer, has the value 1. This
op-code header can be used to identify OpenFlight-format files. By convention, these
files have a “.flt” filename extension.
007-1680-100
253
7: Importing Databases
pfdLoadFile_flt() makes use of several environment variables when locating data and
texture files. These variables and several additional functions, including
pfdConverterMode_flt(), pfdGetConverterMode_flt(), and pfdConverterAttr_flt()
assist in OpenFlight file processing.
McDonnell-Douglas GDS Format
The “.gds” format (also known as the “Things” format) is used in at least one CAD
system, and a minimal loader for this format has been developed for OpenGL Performer
users. The OpenGL Performer loader for “.gds” files is located in the directory
/usr/share/Performer/src/lib/libpfdb/libpfgds for IRIX and Linux and in
%PFROOT%\Src\lib\libpfdb\libpfgds for Microsoft Windows.
The GDS format subset accepted by the pfdLoadFile_gds() function is easy to describe.
It consists of the following five sequential sections in an ASCII file:
1.
The number of vertices, which is given following a “YIN” tag
2. The vertices, with one X, Y, Z triple per line for vertices lines
3. The number zero on a line by itself
4. The number of polygons on a line by itself
5. A series of polygon definitions, each of which is represented on two or more lines.
The first line contains the number one and the name of a material to use for the
polygon. The next line or lines contain the indices for the polygons vertices. The
first number on the first line is the number of vertices. This is followed by that
number of vertex indices on that and possibly subsequent lines.
pfdLoadFile() uses the function pfdLoadFile_gds() to load “.gds” format files into
IRIS Performer.
SGI GFO Format
The GFO format is the simple ASCII format of the barcelona database that is provided
in the OpenGL Performer sample database directory. This database represents the
famous German Pavilion at the Barcelona Exhibition of 1929, which was designed by
Ludwig Mies van der Rohe and is shown in Figure 7-5.
254
007-1680-100
Description of Supported Formats
Figure 7-5
GFO Database of Mies van der Rohe’s German Pavilion
The source code for the GFO-format loader is provided in the file
/usr/share/Performer/src/lib/libpfdb/libpfgfo/pfgfo.c for IRIX and
Linux and in %PFROOT%\Src\lib\libpfdb\libpfgfo\pfgfo.c for Microsoft
Windows.
pfdLoadFile() uses the function pfdLoadFile_gfo() to load GFO format files into
OpenGL Performer run-time data-structures.
When working with GFO files, remember that hardware lighting is not used since all
illumination effects have already been accounted for with the ambient color at each
vertex.
The GFO format defines polygons with a color at every vertex. It is the output format of
an early radiosity system. Files in this format have a simple ASCII structure, as indicated
by the following abbreviated GFO file:
007-1680-100
255
7: Importing Databases
scope {
v3f {42.9632 8.7500 0.9374}
cpack {0x8785a9}
v3f {42.9632 8.0000 0.9374}
cpack {0x8785a9}
...
v3f {-1.0000 -6.5858 10.0000}
cpack {0xffffff}
polygon {cpack[0] v3f[0] cpack[1] v3f[1] cpack[2] v3f[2] cpack[3] v3f[3] }
polygon {cpack[4] v3f[4] cpack[5] v3f[5] cpack[6] v3f[6] cpack[7] v3f[7] }
...
polygon {cpack[7330] v3f[7330] cpack[7331] v3f[7331] cpack[7332] v3f[7332]
cpack[7333] v3f[7333] }
instance {
polygon[0]
polygon[1]
...
polygon[2675]
}
}
This example is taken from the file barcelona-l.gfo, one of only two known
databases in the GFO format. The importer uses functions from the libpfdu library
(such as those from the pfdBuilder) to generate efficient shared triangle strips. This
increases the speed with which GFO databases can be drawn and reduces the size and
complexity of the loader, since the builder’s functions hide the details of the pfGeoSet
construction process.
SGI IM Format
The “.im” format is a simple format developed for test purposes by the OpenGL
Performer engineering team. As new features are added to OpenGL Performer, the “.im”
loader is extended to allow experimentation and testing. A recent example of this is
support for pfText, pfString, and pfFont objects which can be seen by running Perfly on
the sample data file fontsample.im. The OpenGL Performer “.im” loader is in the
directory /usr/share/Performer/src/lib/libpfdb/libpfim for IRIX and
Linux and in %PFROOT%\Src\lib\libpfdb\libpfim for Microsoft Windows.
Here is an example IM format file that creates an extruded 3D text string. Copy this to a
file ending in the extension “.im” and load it into Perfly. For a complete example of how
text is handled in OpenGL Performer, use Perfly to examine the file
256
007-1680-100
Description of Supported Formats
/usr/share/Performer/data/fontsample2.im on IRIX and Linux and in
%PFROOT%\Data\fontsample2.im on Microsoft Windows.
breakup 0 0.0 0 0
new root top
end_root
new font mistr-extruded Mistr 3
end_font
new str_text textnode mistr-extruded 1
Hello World||
end_text
attach top textnode
pfdLoadFile() uses the function pfdLoadFile_im() to load “.im” format files into
OpenGL Performer run-time data structures:
pfdLoadFile_im() searches the current OpenGL Performer file path for the named file
and returns a pointer to the pfNode parenting the imported scene graph, or NULL if the
file is not readable or does not contain a valid database.
AAI/Graphicon IRTP Format
The AAI/Graphicon “.irtp” format is used by the TopGen database modeling system and
by the Graphicon-2000 image generator. The name IRTP is an acronym for Interactive
Real-Time PHIGS. The OpenGL Performer “.irtp” loader is in the directory
/usr/share/Performer/src/lib/libpfdb/libpfirtp for IRIX and Linux and
in %PFROOT%\Src\lib\libpfdb\libpfirtp for Microsoft Windows. Though loader
does not support the more arcane IRTP features, such as binary separating planes or a
global matrix table, it has served as a basis for porting applications to OpenGL Performer
and the RealityEngine.
pfdLoadFile() uses the function pfdLoadFile_irtp() to load IRTP format files into
OpenGL Performer run-time data structures.
007-1680-100
257
7: Importing Databases
SGI Open Inventor Format
The Open Inventor object-oriented 3D-graphics toolkit defines a persistent data format
that is also a superset of the VRML networked graphics data format. The image in
Figure 7-6 shows a sample Open Inventor data file.
Figure 7-6
Aircar Database in IRIS Inventor Format
The model in Figure 7-6 represents one design for the perennial “personal aircar of the
future” concept. It was created, using Imagine, by Mike Halvorson of Impulse, and was
modeled after the Moller 400 as described in Popular Mechanics.
The Open Inventor data-file loader provided with OpenGL Performer reads both binary
and ASCII format Open Inventor data files. Open Inventor scene graph description files
in both formats have the suffix “.iv” appended to their file names.
Here is a simple Open Inventor file that defines a cone:
#Inventor V2.1 ascii
Separator {
Cone {
}
}
258
007-1680-100
Description of Supported Formats
The source code for the Open Inventor format importer is provided in the
libpfdb/libpfiv source directory.
pfdLoadFile() uses the function pfdLoadFile_iv() to load Open Inventor format files into
OpenGL Performer run-time data-structures. OpenGL Performer also comes with an
Inventor loader that works with Open Inventor 2.0, if Open Inventor 2.1 is not installed.
Lightscape Technologies LSA and LSB Formats
The Lightscape Visualization system is a product of Lightscape Technologies, Inc., and is
designed to compute accurate simulations of global illumination within complex 3D
environments. The output files created with Lightscape Visualization can be read into
OpenGL Performer for real-time visual exploration.
Lightscape Technologies provides importers for two of their database formats, the simple
ASCII LSA format and the comprehensive binary LSB format. These loaders are in the
files pflsa.c and pflsb.c in the directories
/usr/share/Performer/src/lib/libpfdb/libpflsa and libpflsb for IRIX
and Linux and in %PFROOT%\Src\lib\libpfdb\libpflsa and libpflsb for
Microsoft Windows. Files in the LSA format are in ASCII and have the following
components:
1.
A 4x4 view matrix representing a default transformation
2. Counts of the number of independent triangles, independent quadrilaterals,
triangle meshes, and quadrilateral meshes in the file
3. Geometric data definitions
There are four types of geometric definitions in LSA files. The formats of these definitions
are as shown in Table 7-7.
Table 7-7
Geometric Definitions in LSA Files
Geometric Type
Format
Triangle
t X1 Y1 Z1 C1 X2 Y2 Z2 C2 X3 Y3 Z3 C3
Triangle mesh
tm n
X1 Y1 Z1 C1
X2 Y2 Z2 C2
...
007-1680-100
259
7: Importing Databases
Geometric Definitions in LSA Files (continued)
Table 7-7
Geometric Type
Format
Quadrilateral
q X1 Y1 Z1 C1 X2 Y2 Z2 C2 X3 Y3 Z3 C3 X4 Y4 Z4 C4
Quadrilateral mesh
qm n
X1 Y1 Z1 C1
X2 Y2 Z2 C2
...
The Cn values in Table 7-7 refer to colors in the format accepted by the OpenGL function
glColor(); these colors should be provided in decimal form. The X, Y, and Z values are
vertex coordinates. Polygon vertex ordering in LSA files is consistently
counterclockwise, and polygon normals are not specified. The first few lines of the LSA
sample file chamber.0.lsa provide an example of the format:
0.486911
-1.665110
0.000000
0.240398
0.03228900
0.00944197
1.92730000
-5.54670000
0.979046
0.286293
-0.017805
13.021200
0.9596590
0.2806240
-0.0174524
13.4945000
1782 4751 0 0
t 4.35 -7.3677 2.57 6188666 6.5 -9.3 2.57 5663353 4.35 -9.3 2.57 5728890
t 6.5 -9.3 2.57 5663353 4.35 -7.3677 2.57 6188666 6.5 -8.2463 2.57 6057596
The count line indicates that the file contains 1782 independent triangles and 4751
independent quadrilaterals, which together represent 11,284 triangles. The image in
Figure 7-7 shows this database, the New Jerusalem City Hall. This was produced by
A.J. Diamond of Donald Schmitt and Company, Toronto, Canada, using the Lightscape
Visualization system.
260
007-1680-100
Description of Supported Formats
Figure 7-7
LSA-Format City Hall Database
pfdLoadFile() uses the function pfdLoadFile_lsa() to load LSA format files into OpenGL
Performer run-time data structures.
Files in the LSB binary format have a very different structure from LSA files.
Representing not just polygon data, they contain much of the structural information
present in the “.ls” files used by the Lightscape Visualization system, including
material, layer, and texture definitions as well as a hierarchical mesh definition for
geometry. This information is structured as a series of data sections, which include the
following:
007-1680-100
•
The signature, a text string that identifies the file
•
The header, which contains global file information
•
The material table, defining material properties
•
The layer table, defining grouping and association
•
The texture table, referencing texture images
•
Geometry in the form of clusters
261
7: Importing Databases
The format of the geometric clusters is somewhat complicated. A cluster is a group of
coplanar surfaces called patches that share a common material, layer, and normal. Each
patch shares at least one edge with another patch in the cluster. Each patch defines either
a convex quadrilateral or a triangle, and patches represent quad-trees called nodes. Each
node points to its corner vertices and its children. The leaf nodes point to their corner
vertices and the child pointers can optionally point to the vertices that split an edge of
the node. Only the locations of vertices that are corners of the patches are stored in the
file; other vertices are created by subdividing nodes of the quad-tree as the LSB file is
loaded. The color information for each vertex is unique and is specified in the file.
The image in Figure 7-8 shows an LSB-format database developed during the design of
a hospital operating room. This database was produced by the DeWolff Partnership of
Rochester, New York, using the Lightscape Visualization system.
Figure 7-8
LSB-Format Operating Room Database
pfdLoadFile() uses the function pfdLoadFile_lsb() to load LSB format files into OpenGL
Performer run-time data structures.
262
007-1680-100
Description of Supported Formats
When working with Lightscape Technologies files, remember that hardware lighting is
not needed because all illumination effects have already been accounted for with the
ambient color at each vertex.
Medit Productions MEDIT Format
The “.medit” format is used by the Medit database modeling system produced by Medit
Productions. The OpenGL Performer “.medit” loader is in the directory
/usr/share/Performer/src/lib/libpfdb/libpfmedit for IRIX and Linux and
in %PFROOT%\Src\lib\libpfdb\libpfmedit for Microsoft Windows.
pfdLoadFile() uses the function pfdLoadFile_medit() to load MEDIT format files into
OpenGL Performer run-time data structures.
NFF Neutral File Format
The “.nff” format was developed by Eric Haines as a way to provide standard procedural
databases for evaluating ray tracing software. OpenGL Performer includes an extended
NFF loader with superquadric torus support, a named build keyword, and numerous
small bug fixes. The “.nff” loader is located in the directory
/usr/share/Performer/src/lib/libpfdb/libpfnff for IRIX and Linux and in
%PFROOT%\Src\lib\libpfdb\libpfnff for Microsoft Windows.
The file /usr/share/Performer/data/sampler.nff on IRIX and Linux and
%PFROOT%\Data\sampler.nff on Microsoft Windows uses each of the NFF data
types. It is an excellent way to explore the “Show Tree”, “Draw Style”, and “Highlight
Mode” features of Perfly. It is included here:
#-- torus
f .75 .00 .25 .6 .8 20 0
t 5 5 0 0 0 1 2 1
build torus
#-- cylinder
f .00 .75 .25 .6 .8 20 0
c
15 5 -3 2
15 5 3 2
#-- put a disc on the top and bottom of the cylinder
d 15 5 -3 0 0 -1 0 2
007-1680-100
263
7: Importing Databases
d 15 5 3 0 0 1 0 2
build cylinder
#-- cone
f .00 .25 .75 .6 .8 20 0
c
25 5 -3 3
25 5 3 0
#-- put a disc on the bottom of the cone
d 25 5 -3 0 0 -1 0 3
build cone
#-- sphere
f .75 .00 .75 .6 .8 20 0
s 5 15 0 3
build sphere
#-- hexahedron
f .25 .25 .50 .6 .8 20 0
h 13 13 -2 17 17 2
build hexahedron
#-- superquadric sphere
f .80 .10 .30 .6 .8 20 0
ss 25 15 0 2 2 2 .1 .4
build superquadric_sphere
#-- disc (washer shape)
f .20 .20 .90 .6 .8 20 0
d 5 25 0 0 0 1 1 2.5
build disc
#-- grid (height field)
f .80 .80 .10 .6 .8 20 0
g 4 4 12 18 22 28 0 4
0 0 0 0
0 1 0 0
0 0 -1 0
0 0 0 0
build grid
#-- superquadric torid
f .40 .20 .60 .6 .8 20 0
st 25 25 0 0.5 0.5 0.5 .33 .33 3
build superquadric_torid
264
007-1680-100
Description of Supported Formats
#-- polygon with no normals
f .20 .20 .20 .6 .8 20 0
p 4
-5 -5 -10
35 -5 -10
35 35 -10
-5 35 -10
build polygon
pfdLoadFile() uses the function pfdLoadFile_nff() to load NFF format files into OpenGL
Performer run-time data structures.
Wavefront Technology OBJ Format
The OBJ format is an ASCII data representation read and written by the Wavefront
Technology Model program. A number of database models in this format have been
placed in the public domain, making this a useful format to have available. OpenGL
Performer provides the function pfdLoadFile_obj() to import OBJ files. The source code
for pfdLoadFile_obj() is in the file pfobj.c in the loader source directory
/usr/share/Performer/src/lib/libpfdb/libpfobj for IRIX and Linux and in
%PFROOT%\Src\lib\libpfdb\libpfobj for Microsoft Windows.
The OBJ-format database shown in Figure 7-9 models an office building that is part of the
SGI corporate campus in Mountain View, California.
007-1680-100
265
7: Importing Databases
Figure 7-9
SGI Office Building as OBJ Database
Files in the OBJ format have a flexible all-ASCII structure, with simple keywords to direct
the parsing of the data. This format is best illustrated with a short example that defines
a texture-mapped square:
#-- ‘v’ defines a vertex; here are four vertices
v -5.000000 5.000000 0.000000
v -5.000000 -5.000000 0.000000
v 5.000000 -5.000000 0.000000
v 5.000000 5.000000 0.000000
#-- ‘vt’ defines a vertex texture coordinate; four are given
vt 0.000000 1.000000 0.000000
vt 0.000000 0.000000 0.000000
vt 1.000000 0.000000 0.000000
vt 1.000000 1.000000 0.000000
#-- ‘usemtl’ means select the material definition defined
#-- by the name MaterialName
usemtl MaterialName
266
007-1680-100
Description of Supported Formats
#-- ‘usemap’ means select the texturing definition defined
#-- by the name TextureName
usemap TextureName
#-- ‘f’ defines a face. This face has four vertices ordered
#-- counterclockwise from the upper left in both geometric
#-- and texture coordinates. Each pair of numbers separated
#-- by a slash indicates vertex and texture indices,
#-- respectively, for a polygon vertex.
f 1/1 2/2 3/3 4/4
pfdLoadFile() uses the function pfdLoadFile_obj() to load Wavefront OBJ files into
OpenGL Performer run-time data structures.
SGI PFB Format
Note: The PFB format is undocumented and is subject to change.
Although OpenGL Performer has no true native database format, the PFB format is
designed to exactly replicate the OpenGL Performer scene graph; this design increases
loading speed. A file in the PFB format has the following advantages:
•
PFB files often load in one tenth (or less) of the time it takes an equivalent file in
another format to load.
•
PFB files are often half the size of equivalent files in another format.
You can think of the PFB format as being a cache. You can convert your files into PFB for
fast and efficient loading or paging, but you should always keep your original files in
case you wish to modify them.
Converting to the PFB Format
You can convert files into the PFB format in one of the following ways:
007-1680-100
•
Use the function pfdStoreFile_pfb() in libpfpfb.
•
Use pfconv.
267
7: Importing Databases
SGI PFI Format
The PFI image file format is designed for fast loading of images into pfTextures.
pfLoadTexFile() can load PFI files as the image of a pfTexture. Since the format of the
image in a PFI file matches that of a pfTexture, data is not reformatted at load time.
Eliminating the reformatting often cuts the load time of textures to half of the load time
of the same image in the IRIS RGB image format.
PFI files can contain the mipmaps of the image. This feature saves significant time in the
OpenGL Performer DRAW process since it does not have to generate the mipmaps.
Creating PFI Files
PFI files are created in the following ways:
•
pfSaveTexFile() creates a PFI file from a pfTexture.
•
The pfdImage methods in libpfdu create PFI files.
•
pficonv converts IRIS RGB image files into PFI files.
•
pfconv converts all referenced image files into PFI files when the setting
PFPFB_SAVE_TEXTURE_PFI mode is PF_ON. The command line options to do this
with pfconv is -Mpfb,5.
SGI PHD Format
The PHD format was created to describe the geometric polyhedron definitions derived
mathematically by Andrew Hume and by the Kaleido program of Zvi Har’El. This
format describes only the geometric shape of polyhedra; it provides no specification for
color, texture, or appearance attributes such as specularity.
The OpenGL Performer sample data directories contain numerous polyhedra in the PHD
format. The image in Figure 7-10 shows many of the polyhedron definitions laboriously
computed by Andrew Hume.
268
007-1680-100
Description of Supported Formats
Figure 7-10
Plethora of Polyhedra in PHD Format
The source code for the PHD-format importer is in the file
/usr/share/Performer/src/lib/libpfdb/libpfpoly/pfphd.c on IRIX and
Linux and in %PFROOT%\Src\lib\libpfdb\libpfpoly\pfdhd.c on Microsoft
Windows.
PHD format files have a line-structured ASCII form; an initial keyword defines the
contents of each line of data. The file format consists of a filename definition (introduced
by the keyword file) followed by one or more object definitions.
Object definitions are bracketed by the keywords object.begin and object.end and
contain one or more polygon definitions. Objects can have a name in quotes following
the object.begin keyword; such a name is used by the loader for the name of the
corresponding OpenGL Performer node.
Polygon definitions are bracketed by the keywords polygon.begin and
polygon.end and contain three or more vertex definitions.
007-1680-100
269
7: Importing Databases
Vertex definitions are introduced by the vertex keyword and define the X, Y, and Z
coordinates of a single vertex.
The following is a PHD-format definition of a unit-radius tetrahedron centered at the
origin of the coordinate axes. It is derived from the database developed by Andrew
Hume but has since been translated, scaled, and reformatted.
file 000.phd
object.begin "tetrahedron"
polygon.begin
vertex -0.090722 -0.366647
vertex 0.544331 -0.628540
vertex 0.453608 0.890430
polygon.end
polygon.begin
vertex -0.907218 0.104757
vertex -0.090722 -0.366647
vertex 0.453608 0.890430
polygon.end
polygon.begin
vertex -0.090722 -0.366647
vertex -0.907218 0.104757
vertex 0.544331 -0.628540
polygon.end
polygon.begin
vertex 0.453608 0.890430
vertex 0.544331 -0.628540
vertex -0.907218 0.104757
polygon.end
object.end
0.925925
-0.555555
0.037037
-0.407407
0.925925
0.037037
0.925925
-0.407407
-0.555555
0.037037
-0.555555
-0.407407
pfdLoadFile() uses the function pfdLoadFile_phd() to load PHD format files into
OpenGL Performer run-time data structures.
The pfdLoadFile_phd() function composes a color with red, green, and blue components
uniformly distributed within the range 0.2 to 0.7 that is consistent for each polygon with
the same number of vertices within a single polyhedron.
SGI PTU Format
The PTU format is named for the OpenGL Performer Terrain Utilities, of which the
pfdLoadFile_ptu() function is the sole example at the present time. This function accepts
270
007-1680-100
Description of Supported Formats
as input the name of a control file (the file with the “.ptu” filename extension) that
defines the desired terrain parameters and references additional data files.
The database shown in Figure 7-11 represents a portion of the Yellowstone National Park.
This terrain database was generated completely by the OpenGL Performer Terrain
Utility data generator from digital terrain elevation data and satellite photographic
images. Image manipulation is performed using the SGI ImageVision Library functions.
Figure 7-11
Terrain Database Generated by PTU Tools
The PTU control file has a fixed format that does not use keywords. The contents of this
file are simply ASCII values representing the following data items:
1.
The name to be assigned to the top-level pfNode built by pfdLoadFile_ptu().
2. The number of desired levels-of-detail (LOD) for the resulting terrain surface. The
pfdLoadFile_ptu() function will construct this many versions of the terrain, each
representing the whole surface but with exponentially fewer numbers of polygons
in each version.
007-1680-100
271
7: Importing Databases
3. The number of highest-LOD tiles that will tessellate the entire terrain surface in the
X and Y axis directions.
4. Two numeric values that define the mapping of texture image pixels to
world-coordinate terrain geometry. These values are the number of meters per texel
(texture pixel) of filtered grid post data in the X and Y axis dimensions.
5. The name of an image file that represents terrain height at regularly spaced sample
points in the form of a monochrome image whose brightness at each pixel indicates
the height at that sample point. Additional arguments are the number of samples in
the input image in the X and Y directions, as well as the desired number of samples
in these directions. The pfdLoadFile_ptu() function resamples the grid posts from
the original to the desired resolution by filtering the height image using SGI
ImageVision Library functions.
6. The name of an image file that represents the terrain texture image at regularly
spaced sample points. Subsequent arguments are the number of samples in the
image in the X and Y directions as well as the desired number of samples in these
directions. This image will be applied to the terrain geometry. The scale values
provided in the PTU file allow the terrain grid and texture image to be adjusted to
create an orthographic alignment.
7. An optional second texture-image filename that serves as a detail texture when the
terrain is viewed on RealityEngine systems. This texture is used in addition to the
base texture image.
8. An optional detail-texture spline-table definition. The blending of the primary
texture image and the secondary detail texture is controlled by a blend table defined
by this spline function. The spline table is optional even when a detail texture is
specified. Detail texture and its associated blend functions are applicable only on
RealityEngine systems.
The source code for the PTU-format importer is provided in the file
/usr/share/Performer/src/lib/libpfdb/libpfptu/pfptu.c on IRIX and
Linux and in %PFROOT%\Src\lib\libpfdb\libpfptu\pfptu.c on Microsoft
Windows.
pfdLoadFile() uses the function pfdLoadFile_ptu() to load PTU format files into
OpenGL Performer run-time data structures.
272
007-1680-100
Description of Supported Formats
ArchVision RPC Format
ArchVision provides the rich photorealistic content (RPC) loader. The RPC loader loads
in images from an ArchVision RPC file. The images represent views of an object from a
set of directions around the object. If you provide an existing pfIBRnode, the images are
loaded into a pfIBRtexture of the node. Otherwise, the function creates a new pfIBRnode
with a single pfGeoSet and a pfIBRtexture containing the images. In the case of new
content with meshes, the pfGeoSet contains the mesh, which becomes the proxy in the
pfIBRnode.
The following functions allow you to access and alter the modes, values, and attributes
of the RPC loader:
•
pfdConverterMode_rpc(), pfdGetConverterMode_rpc()
•
pfdConverterVal_rpc(), pfdGetConverterVal_rpc()
•
pfdConverterAttr_rpc(), pfdGetConverterAttr_rpc()
You control the RPC converter modes with the token PFRPC_USE_USER_IBRNODE. By
default, the loader creates a pfIBRnode with a single pfGeoSet and a pfIBRtexture that
contains the loaded images. If this mode is set to PF_ON and you supply a pfIBRnode
using pfdConverterAttr_rpc(), the images are loaded into the pfIBRtexture of that node.
Table 7-8 describes the RPC converter values.
Table 7-8
007-1680-100
RPC Converter Values
Converter Value
Description
PFRPC_SKIP_TEXTURES
Skips every n images. ArchVision RPC files often
contain hundreds of images. A pfIBRtexture containing
so many images would be too large. The default is set
to 2. If you want to use all images in the file, set it to 0.
PFRPC_CROP_LEFT
Crops the loaded images by the specified number of
pixels on the left.
PFRPC_CROP_RIGHT
Crops the loaded images by the specified number of
pixels on the right. Note that the resulting image width
should be a power of 2.
PFRPC_CROP_TOP
Crops the loaded images by the specified number of
pixels on the top.
273
7: Importing Databases
Table 7-8
274
RPC Converter Values (continued)
Converter Value
Description
PFRPC_CROP_BOTTOM
Crops the loaded images by the specified number of
pixels on the bottom. Note that the resulting image
height should be a power of 2.
PFRPC_SCALE_WIDTH
Scales the billboard width in the case of a pfIBRnode
without a proxy.
PFRPC_SCALE_HEIGHT
Scales the billboard height in the case of a pfIBRnode
without a proxy.
PFRPC_NEAREST
Sets flag PFIBR_NEAREST on the pfIBRnode created
by the loader.
PFRPC_USE_NEAREST_RING
In the case of content with a proxy and having more
than one ring of views, forces the mode in which views
are selected from the nearest ring rather than having
the views blended between the two nearest rings.
PFRPC_COMBINED_TEXTURE_SIZE
Combines textures into a square texture of the specified
size (should be a power of 2). By default, if the texture
size is not a power of 2, textures are combined into a
texture of size 2048x2048. If the texture size is power of
2, textures are not combined into a bigger texture unless
the value PFRPC_COMBINED_TEXTURE_SIZE is
explicitly specified. You can also set it to 0 to disable
combining.
007-1680-100
Description of Supported Formats
Table 7-9 describes the converter attributes.
Table 7-9
RPC Converter Attributes
Converter Attribute
Description
PFRPC_USER_IBRNODE
Specifies a pfIBRnode. The images from the RPC file are
loaded into the pfIBRtexture of the node.
PFRPC_RING_FILE
Specifies the path to ring files that define the rings of
views where proxies are used. There is one file for each
component of the input RPC file, indexed by extension
.0, .1, .2, and so on. Each line of the ring file contains the
angle of the ring from the horizon and the number of
views in that ring. If no ring file is specified, each
component has only one ring of 16 views at horizontal
angle 0. You can use an environment variable of the
same name to set this attribute.
PFRPC_SKIP_TEXTURES
PFRPC_SCALE_WIDTH
PFRPC_FLIP_TEXTURES
PFRPC_NEAREST
PFRPC_USE_NEAREST_RING
PFRPC_COMBINED_TEXTURE_SIZE
See Table 7-8 for the descriptions of these attributes. You
can use an environment variable of the same name to set
this attribute. Setting attribute values through the use of
environment variables allows you to affect the loading
of the files without the necessity of changing your
application.
Note: The loader is using a relatively slow, third-party routine for decompressing
images. For a faster load time, you may want to convert your RPC files into PFB files
using pfconv.
Two sample RPC files can be found in directory
/usr/share/Performer/data/ibr/rpc for IRIX and Linux and in
%PFROOT%\Data\ibr\rpc for MicroSoft Windows. You can download other files from
the ArchVision webpage at www.archvision.com.
USNA Standard Graphics Format
The SGF format is used at the United States Naval Academy as a standard graphics
format for geometric data. The loader was developed based on the description of the
007-1680-100
275
7: Importing Databases
standard graphics format as described by David F. Rogers and J. Alan Adams in the book
Mathematical Elements for Computer Graphics. The OpenGL Performer “.sgf” format loader
is located in the directory /usr/share/Performer/src/lib/libpfdb/libpfsgf
for IRIX and Linux and in %PFROOT%\Src\lib\libpfdb\libpfsgf for Microsoft
Windows
Here is the vector definition for four stacked squares in SGF form:
0, 0, 0
1, 0, 0
1, 1, 0
0, 1, 0
0, 0, 0
1.0e37,
0, 0, 1
1, 0, 1
1, 1, 1
0, 1, 1
0, 0, 1
1.0e37,
0, 0, 2
1, 0, 2
1, 1, 2
0, 1, 2
0, 0, 2
1.0e37,
0, 0, 3
1, 0, 3
1, 1, 3
0, 1, 3
0, 0, 3
1.0e37,
1.0e37, 1.0e37
1.0e37, 1.0e37
1.0e37, 1.0e37
1.0e37, 1.0e37
pfdLoadFile() uses the function pfdLoadFile_sgf() to load SGF format files into OpenGL
Performer run-time data-structures.
SGI SGO Format
The SGI Object format is used by several utility programs and was one of the first
database formats supported by OpenGL Performer. The image in Figure 7-12 shows a
model generated by Paul Haeberli and loaded into Perfly by the pfdLoadFile_sgo()
database importer.
276
007-1680-100
Description of Supported Formats
Figure 7-12
Model in SGO Format
Objects in the SGO format have per-vertex color specification and multiple data formats.
Objects contained in SGO files are constructed from three data types:
•
Lists of quadrilaterals
•
Lists of triangles
•
Triangle meshes
Objects of different types can be included as data within one SGO file.
The SGO format has the following structure:
1.
A magic number, 0x5424, which identifies the file as an SGO file.
2. A set of data for each object. Each object definition begins with an identifying token,
followed by geometric data. There can be multiple object definitions in a single file.
An end-of-data token terminates the file.
007-1680-100
277
7: Importing Databases
The layout of an SGO file is the following:
<SGO-file magic number>
<data-type token for object #1>
<data for object #1>
<data-type token for object #2>
<data for object #2>
...
<data-type token for object #n>
<data for object #n>
<end-of-data token>
Each of the identifying tokens is 4 bytes long. Table 7-10 lists the symbol, value, and
meaning for each token.
Table 7-10
Object Tokens in the SGO Format
Symbol
Value
Meaning
OBJ_QUADLIST
1
Independent quadrilaterals
OBJ_TRILIST
2
Independent triangles
OBJ_TRIMESH
3
Triangle mesh
OBJ_END
4
End-of-data token
The next word following any of the three object types is the number of 4-byte words of
data for that object. The format of this data varies depending on the object type.
For quadrilateral list (OBJ_QUADLIST) and triangle list (OBJ_TRILIST) objects, there are
nine words of floating-point data for each vertex, as follows:
1.
Three words that specify the components of the normal vector at the vertex
2. Three words that specify the red, green, and blue color components, scaled to the
range 0.0 to 1.0
3. Three words that specify the X, Y, and Z coordinates of the vertex itself
In quadrilateral lists, vertices are in groups of four; so, there are 4 × 9 = 36 words of data
for each quadrilateral. In triangle lists, vertices are in groups of three, for 3 x 9 = 27 words
per triangle.
278
007-1680-100
Description of Supported Formats
The triangle mesh, OBJ_TRIMESH, is the most complicated of the three object data types.
Triangle mesh data consists of a set of vertices followed by a set of mesh-control
commands. Triangle mesh data has the following format:
1.
A long word that contains the number of words in the complete triangle mesh data
packet
2. A long word that contains the number of floating-point words required by the
vertex data, at nine words per vertex
3. The data for each vertex, consisting of nine floating-point words representing
normal, color, and coordinate data
4. A list of triangle mesh controls
The triangle mesh controls, each of which is one word in length, are listed in Table 7-11.
Table 7-11
Mesh Control Tokens in the SGO Format
Symbol
Value
Meaning
OP_BGNTMESH
1
Begin a triangle strip.
OP_SWAPTMESH
2
Exchange old vertices.
OP_ENDBGNTMESH
3
End then begin a strip.
OP_ENDTMESH
4
Terminate triangle mesh.
The triangle-mesh controls are interpreted sequentially. The first control must always be
OP_BGNTMESH, which initiates the mesh-decoding logic. After each mesh control is a
word (of type long integer) that indicates how many vertex indices follow. The vertex
indices are in byte offsets, so to access vertex n, you must use the byte offset n x 9 x 4.
pfdLoadFile() uses the function pfdLoadFile_sgo() to load SGO format files into
OpenGL Performer run-time data structures.
You can find the source code for the SGO-format importer in the file pfsgo.c. This
importer does not attempt to decode any triangle meshes present in input files; instead,
it terminates the file conversion process as soon as an OBJ_TRIMESH data-type token is
encountered. If you use SGO-format files containing triangle meshes you will need to
extend the conversion support to include the triangle mesh data type.
007-1680-100
279
7: Importing Databases
USNA Simple Polygon File Format
The “.spf” format is used at the United States Naval Academy as a simple polygon file
format for geometric data. The loader was developed based on the description in the
book Mathematical Elements for Computer Graphics. The OpenGL Performer “.spf” loader
is in the directory /usr/share/Performer/src/lib/libpfdb/libpfspf on IRIX
and Linux and in %PFROOT%\Src\lib\libpfdb\libpfspf on Microsoft Windows.
The following “.spf” format file is defined in that book.
polygon with a hole
14,2
4,4
4,26
20,26
28,18
28,4
21,4
21,8
10,8
10,4
10,12
10,20
17,20
21,16
21,12
9,1,2,3,4,5,6,7,8,9
5,10,11,12,13,14
If you look at this file in Perfly, you will see that the hole is not cut out of the letter “A”
as might be desired. Such computational geometry computations are not considered the
province of simple database loaders.
pfdLoadFile() uses the function pfdLoadFile_spf() to load SPF format files into
OpenGL Performer run-time data structures.
Sierpinski Sponge Loader
The Sierpinski Sponge (also known as Menger Sponge) loader is not based on a data
format but rather is a procedural data generator. The loader interprets the portion of the
user-provided filename before the period and extension as an integer which specifies the
number of recursive subdivisions desired in data generation. For example, providing the
280
007-1680-100
Description of Supported Formats
pseudo filename “3.sponge” to Perfly will result in the Sponge loader being invoked and
generating a sponge object using three levels of recursion, resulting in a 35712 polygon
database object. The OpenGL Performer “.sponge” loader can be found in the directory
/usr/share/Performer/src/lib/libpfdb/libpfsponge on IRIX and Linux
and in %PFROOT%\Src\lib\libpfdb\libpfsponge on Microsoft Windows.
pfdLoadFile() uses the function pfdLoadFile_sponge() to load Sponge format files into
OpenGL Performer run-time data structures.
Star Chart Format
The “.star” format is a distillation of astronomical data from the Yale Compact Star Chart
(YCSC). The sample data file /usr/share/Performer/data/3010.star for IRIX
and Linux and %PFROOT%\Data\3010.star for Microsoft Windows contains data
from the YCSC that has been reduced to a list of the 3010 brightest stars as seen from
Earth and positioned as 3010 points of light on a unit-radius sphere. The OpenGL
Performer “.star” loader can read this data and is provided as a convenience for making
dusk, dawn, and night-time scenes. The loader is in the directory
/usr/share/Performer/src/lib/libpfdb/libpfstar on IRIX and Linux and
in %PFROOT%\Src\lib\libpfdb\libpfstar on Microsoft Windows.
Data in a “.star” file is simply a series of ASCII lines with the “s” (for star) keyword
followed by X, Y, and Z coordinates, brightness, and an optional name. Here are the 10
brightest stars (excluding Sol) in the “.star” format:
s
s
s
s
s
s
s
s
s
s
-0.18746032 0.93921369 -0.28763914 1.00 Sirius
-0.06323564 0.60291260 -0.79529721 1.00 Canopus
-0.78377002 -0.52700269 0.32859191 1.00 Arcturus
0.18718566 0.73014212 0.65715599 1.00 Capella
0.12507832 -0.76942003 0.62637711 0.99 Vega
0.13051330 0.68228769 0.71933979 0.99 Capella
0.19507207 0.97036278 -0.14262892 0.98 Rigel
-0.37387931 -0.31261155 -0.87320572 0.94 Rigil Kentaurus
-0.41809806 0.90381104 0.09121194 0.94 Procyon
0.49255905 0.22369388 -0.84103900 0.92 Achernar
pfdLoadFile() uses the function pfdLoadFile_star() to load Star format files into
OpenGL Performer run-time data structures.
007-1680-100
281
7: Importing Databases
3D Lithography STL Format
The STL format is used to define 3D solids to be imaged by 3D lithography systems. STL
defines objects as collections of triangular facets, each with an associated face normal.
The ASCII version of this format is known as STLA and has a very simple structure.
The image in Figure 7-13 shows a typical STLA mechanical CAD database. This model is
defined in the bendix.stla sample data file.
Figure 7-13
Sample STLA Database
The source code for the STLA-format loader is in the files
/usr/share/Performer/src/lib/libpfdb/libpfstla/pfstla.c on IRIX and
Linux and in %PFROOT%\Src\lib\libpfdb\libpfstla\pfstla.c on Microsoft
Windows.
STLA-format files have a line-structured ASCII form; an initial keyword defines the
contents of each line of data. An STLA file consists of one or more facet definitions, each
of which contains the following:
1.
The facet normal, indicated with the facet normal keyword
2. The facet vertices, bracketed by outer loop and endloop keywords
282
007-1680-100
Description of Supported Formats
3. The endloop keyword
Here is an excerpt from nut.stla, one of the STLA files provided in the OpenGL
Performer sample data directories. These are the first two polygons of a 524-triangle
hex-nut object:
facet normal 0 -1 0
outer loop
vertex 0.180666 -7.62 2.70757
vertex -4.78652 -7.62 1.76185
vertex -4.436 -7.62 0
endloop
endfacet
facet normal -0.381579 -0.921214 -0.075915
outer loop
vertex -4.48833 -7.59833 0
vertex -4.436 -7.62 0
vertex -4.78652 -7.62 1.76185
endloop
endfacet
Use this function to import data from STLA-format files into OpenGL Performer
run-time data structures:
pfNode *pfdLoadFile_stla(char *fileName);
pfdLoadFile_stla() searches the current OpenGL Performer file path for the file named
by the fileName argument and returns a pointer to the pfNode that parents the imported
scene graph, or NULL if the file is not readable or does not contain recognizable STLA
format data.
SuperViewer SV Format
The SuperViewer (SV) format is one of the several database formats that the I3DM
database modeling tool can read and write. The I3DM modeler was developed by John
Kichury of SGI and is provided with OpenGL Performer. The source code for the SV
format importer is in the file pfsv.c.
The passenger vehicle database shown in Figure 7-14 was modeled using I3DM and is
stored in the SV database format.
007-1680-100
283
7: Importing Databases
Figure 7-14
Early Automobile in SuperViewer SV Format
Within SV files, object geometry and attributes are described between text lines that
contain the keywords model and endmodel. For example:
model wing
geometry and attributes
endmodel
Any number of models can appear within a SuperViewer file. The geometry and
attribute data mentioned above each consist of one of the following types:
•
3D Polygon with vertex normals and optional texture coordinates:
poly3dn <num_vertices> [textured]
x1 y1 z1 nx1 ny1 nz1 [s1 t1]
x2 y2 z2 nx2 ny2 nz2 [s2 t2]
...
where the coordinates and normals are defined as follows:
284
–
Xn Yn Zn are the nth vertex coordinates
–
Nxn Nyn Nzn are the nth vertex normals
007-1680-100
Description of Supported Formats
–
•
Sn Tn are the nth texture coordinates
3D Triangle mesh with vertex normals and optional texture coordinates
tmeshn <num_vertices> [textured]
x1 y1 z1 nx1 ny1 nz1 [s1 t1]
x2 y2 z2 nx2 ny2 nz2 [s2 t2]
...
where the coordinates and normals are defined as follows:
•
–
Xn Yn Zn are the nth vertex coordinates
–
Nxn Nyn Nzn are the nth vertex normals
–
Sn Tn are the nth texture coordinates
Material definition. If the material directive exists before a model definition, it is
taken as a new material specification. Its format is the following:
material n Ar Ag Ab Dr Dg Db Sr Sg Sb Shine Er Eg Eb
where the variables represent the following:
–
n is an integer specifying a material number
–
Ar Ag Ab is the ambient color.
–
Dr Dg Db is the diffuse color.
–
Sr Sg Sb is the specular color.
–
Shine is the material shininess.
–
Er Eg Eb is the emissive color.
If the material directive exists within a model description, the format is the
following:
material n
where n is an integer specifying which material (as defined by the material
description above) is to be assigned to subsequent data.
•
Texture definition. If the texture directive exists before a model definition it is taken
as a new texture specification. Its format is the following:
texture n TextureFileName
If the texture directive exists within a model description, the format is the following:
texture n
007-1680-100
285
7: Importing Databases
where n is an integer specifying which texture (as defined by the texture description
above) is to be assigned to subsequent data.
•
Backface polygon display mode. The backface directive is specified within model
definitions to control backface polygon culling:
backface mode
where a mode of “on” allows the display of backfacing polygons and a mode of “off”
suppresses their display.
In actual use the SV format is somewhat self-documenting. Here is part of the SV file
apple.sv from the directory /usr/share/Performer/data on IRIX and Linux and
in %PFROOT%\Data on Microsoft Windows:
material 20 0.0 0.0 0 0.400000 0.000000 0 0.333333 0.000000 0.0 10.0000 0 0 0
material 42 0.2 0.2 0 0.666667 0.666667 0 0.800000 0.800000 0.8 94.1606 0 0 0
material 44 0.0 0.2 0 0.000000 0.200000 0 0.000000 0.266667 0.0 5.0000 0 0 0
texture 4 prchmnt.rgb
texture 6 wood.rgb
model LEAF
material 44
texture 4
backface on
poly3dn 4 textured
1.35265 1.35761 13.8338
0.88243 0.96366 14.0329
-4.44467 1.24026 13.5669
-2.37938 2.17479 13.3626
poly3dn 4 textured
-2.37938 2.17479 13.3626
-4.44467 1.24026 13.5669
-9.23775 2.34664 13.1475
-6.69592 3.94535 12.6716
0.0686595
0.0502096
0.0363863
0.0363863
-0.234553
-0.376701
-0.337291
-0.337291
-0.969676
-0.924973
-0.940697
-0.940697
0 1
0 0.75
0.0909091 0.75
0.0909091 1
0.0363863
0.0363863
0.0344832
0.0344832
-0.337291
-0.337291
-0.284369
-0.284369
-0.940697
-0.940697
-0.958095
-0.958095
0.0909091 1
0.0909091 0.75
0.181818 0.75
0.181818 1
This excerpt specifies material properties and references texture images stored in the files
prchmnt.rgb and wood.rgb, and then defines two polygons.
pfdLoadFile() uses the function pfdLoadFile_sv() to load SuperViewer files into
OpenGL Performer run-time data structures.
286
007-1680-100
Description of Supported Formats
Geometry Center Triangle Format
The “.tri” format is used at the University of Minnesota’s Geometry Center as a simple
geometric data representation. The loader was developed by inspection of a few sample
files. The OpenGL Performer “.tri” loader is in the directory
/usr/share/Performer/src/lib/libpfdb/libpftri on IRIX and Linux and in
%PFROOT%\Src\lib\libpfdb\libpftri on Microsoft Windows.
These files have a very simple format: a line per vertex with position and normal given
on each line as 6 ASCII numeric values. The file is simply a series of these triangle
definitions. Here are the first two triangles from the data file
/usr/share/Performer/data/mobrect.tri on IRIX and Linux and in
%PFROOT%\Data\mobrect.tri on Microsoft Windows:
1.788180 1.000870 0.135214 0.076169 -0.085488 0.993423
1.574000 0.925908 0.146652 0.089015 -0.086072 0.992304
1.793360 0.634711 0.099409 0.076402 -0.111845 0.990784
0.836848 -0.595230 0.197960 0.156677 0.044503 0.986647
0.709638 -0.345676 0.210010 0.157642 0.021968 0.987252
0.581200 -0.535321 0.234807 0.145068 0.030985 0.988936
pfdLoadFile() uses the function pfdLoadFile_tri() to load “.tri” format files into OpenGL
Performer run-time data structures.
UNC Walkthrough Format
The “.unc” format was once used at the University of North Carolina as a format for
geometric data in an architectural walkthrough application. The loader was developed
based on inspection of a few sample files. The OpenGL Performer “.unc” loader is in the
directory /usr/share/Performer/src/lib/libpfdb/libpfunc for IRIX and
Linux and in %PFROOT%\Src\lib\libpfdb\libpfunc for Microsoft Windows.
pfdLoadFile() uses the function pfdLoadFile_unc() to load UNC format files into
OpenGL Performer run-time data structures.
WRL Format
The VRML 2.0 format for OpenGL Performer, wrl, is made by DRaW Computing
Associates. It accepts geometry and texture only. Basic geometry nodes like Sphere,
Cone, Cylinder, Box and related nodes like Shape, Material, Appearance,
007-1680-100
287
7: Importing Databases
TextureTransform, ImageTexture, and ElevationGrid are supported. Also, complex
geometries can be obtained using the IndexedFaceSet node. You can do geometric
manipulations to nodes using Group nodes and Transform nodes. You can also make
very complex structures using PROTOs, where you group many geometry nodes.
Database Operators with Pseudo Loaders
The OpenGL Performer dynamic database loading mechanism provides additional
DSOs that operate on the resulting scene graph from a file or set of files after the file(s)
are loaded. This mechanism, called “pseudo loaders,” enables the desired operator DSO
to be specified as additional suffixes to the filename. The DSO matching the last suffix is
loaded first and provided the entire filename. That pseudo loader then can parse the
arbitrary filename and invoke the next operator or loader and then operate on the results.
This process allows additional arguments to be buried in the specified filename for the
pseudo loader to detect and parse.
One set of pseudo loaders included with OpenGL Performer are the rot, trans, and
scale loaders. These loaders take hpr and xyz arguments in addition to their Filename
and can be invoked from any program using pfdLoadFile(), as shown in this example:
% perfly cow.obj.-90,90,0.rot
-90, 90, and 0 are the h, p, and r values, respectively.
If you are using a shell with argument expansion, such as csh, you can create interesting
cow art. Try out the following example:
% perfly cow.obj.{0,1},0,0.trans cow.obj.{0,1,2,3,4},0,-5.trans
Specifying a base filename is only needed if the specified pseudo loader expects a file to
process. Loaders can generate their scene graphics procedurally based on simple
parameters specified in the command string.
288
007-1680-100
Database Operators with Pseudo Loaders
The pseudo loaders in the OpenGL Performer distribution are described in Table 7-12.
Table 7-12
OpenGL Performer Pseudo Loaders
Pseudo Loaders
Description
libpfrot
Add pfSCS at root to rotate scene graph by specified h,p,r.
libpftrans
Add pfSCS at root to translate scene graph by specified x,y,z.
libpfscale
Add pfSCS at root to sale scene graph by specified x,y,z.
libpfclosest
Adds run-time application callback to highlight closest point each frame.
libpfcliptile
Adds callback to compute for the specified tilename, minS ,minT, maxS, and
maxT, the proper virtual cliptexture viewing parameters.
libpfsphere
Generates a sphere database with morphing LOD starting from an n-gon for
specified n, power of 2.
libpfvct
Convert normal cliptexture .ct file to virtual cliptexture.
libpfsubdiv
Subdivide an arbitrary file using Loop or Catmull-Clark subdivision.
libpfgeoa
Convert geometry from pfGeoSets to pfGeoArrays.
libpfgopt
Optimize the geometry of pfGeoSets or pfGeoArrays.
libpfbreakup
Create an artificial hierarchy from unstructured input.
Pseudo loaders should define pfdLoadNeededDSOs_EXT() for the following:
•
Preinitializing DSOs
•
Loading other special files
•
Performing additional initialization, such as class initialization, that should happen
before pfConfig()
The libpfbreakup pseudo loader, which creates an artificial hierarchy from
unstructured input, uses the pfdBreakup() function, whose syntax follows:
pfNode * pfdBreakup(pfGeode *geode, float geodeSize, int stripLength, int geodeChild);
The function accepts a pfGeode that contains pfGeoSets of type PFGS_TRISTRIPS and
builds a new scene graph with the same geometric content but with a spatial subdivision
structure designed for efficient processing. The function returns the root of the new scene
graph (a pfGroup) or, if the subdivision was not done, a copy of the original pfGeode.
007-1680-100
289
7: Importing Databases
The first triangle strips of all pfGeoSets in geode are split into strips no longer than
stripLength. If the pfGeoSets do not contain triangle strips, they are left untouched
and geode is subdivided based on the geometrical centers of the split pfGeoSets using
an octree. The degree of recursive partitioning desired is specified in the function
arguments. The resulting scene graph is a pfGroup that contains more pfGroups,
recursively. The recursion stops if the resulting pfGeode is smaller than geodeSize
(geodeSize is the maximum size of the leaf octants in world coordinates) or the number
of its pfGeoSets is smaller than geodeChild.
Note: The input pfGeode is not deleted.
See also “The libpfsubdiv Pseudo Loader” on page 424.
The Maya Database Exporter
OpenGL Performer provides the PFBexport plug-in for Maya. This Maya exporter
converts a Maya scene into OpenGL Performer data structures and saves them in an
OpenGL Performer binary (.pfb) or ASCII (.pfa) file. These output files can be
displayed with the OpenGL Performer viewer or imported into an OpenGL Performer
application. The exporter produces a log file that describes the Maya objects converted
and flags any errors or unsupported features.
This section describes the following topics:
•
“Installation Requirements”
•
“Exporting a Scene Using the Graphical Interface”
•
“Exporting a Scene Using the Maya Embedded Language (MEL)”
•
“Translation Details”
Installation Requirements
The Maya exporter should be installed for you automatically as part of the OpenGL
Performer installation process. You must have a licensed copy of Maya 4.5 or later
installed on every machine that runs the exporter because it is a Maya plug-in and can
only run within the Maya environment. This differs from previous OpenGL Performer
290
007-1680-100
The Maya Database Exporter
file loaders that run within the OpenGL Performer environment. The Maya exporter is
available for both IRIX and Windows, but not for Linux.
The .pfa and .pfb files produced by the Maya exporter can be viewed on any machine
with OpenGL Performer 3.1 or later installed. These files are not compatible with
previous versions of OpenGL Performer. Maya has capabilities like subdivision surfaces
and non-uniform rational B-splines (NURBS) that require the OpenGL Performer 3.1 or
later run-time environment. The exporter can optimize geometry to take advantage of
the new OpenGL extensions available on Onyx4 systems. These extensions also need
OpenGL Performer 3.1 or later support.
If you launch Maya and it cannot find the plug-in, you can troubleshoot by checking the
MAYA_PLUG_IN_PATH and MAYA_SCRIPT_PATH environment variables. Both
variables should be defined at installation to reference the OpenGL Performer directory
containing the Maya extensions.
Table 7-13 shows the default path for IRIX and Microsoft Windows.
Table 7-13
Default Path for the Maya Export Plug-in
Platform
Default Path
IRIX
/usr/share/OpenGL Performer/bin
Microsoft
Windows
c:\SGI\OpenGL Performer\Bin
Exporting a Scene Using the Graphical Interface
Any Maya scene can be exported to OpenGL Performer format. Unsupported features
will be flagged in the export log file and may result in objects missing from the scene
when viewed with OpenGL Performer. To run the plug-in from the Maya graphical user
interface, use the Export All or Export Selection items from the File menu. As shown in
Figure 7-15, you should be able to select PFBexport as one of the file types. The only
supported extensions are .pfb and .pfa.
007-1680-100
291
7: Importing Databases
Figure 7-15
Maya Export Screen
There are a number of parameters that control how Maya content is translated to
OpenGL Performer format. You can produce exported files that take advantage of
OpenGL extensions for multitexturing and vertex buffer objects but you must be running
Maya on a platform which has these features. In general, OpenGL Performer .pfa and
.pfb files are cross-platform and can be successfully imported on machines that are
less-capable than where they originated with some loss of fidelity. For example,
displaying a multitextured object on an Octane will drop all textures but the first.
As shown in Figure 7-16, you can optimize for earlier versions of SGI hardware by
choosing Optimize for OpenGL 1.0. If you have have an Onyx 4 system, do not check this
option or you will not get the best graphics performance. If your graphics hardware does
not have multitexturing capabilities, do check this option.
292
007-1680-100
The Maya Database Exporter
Figure 7-16
007-1680-100
Maya Export Options
293
7: Importing Databases
Table 7-14 describes the export options.
Table 7-14
Maya Export Options
Category
Option
Description
Export Options
Export geometry only
If checked, only geometry will be included
in the are also exported. This can result in
poor performance from using too many
light sources.
Optimize for OpenGL 1.0
If checked, the exporter optimizes for earlier
versions of SGI hardware without
multitexturing capabilities. Do not check
this option for an Onyx4 system.
Visual preview
If checked, you get a visual preview.
Indexed meshes
If checked, polygon mesh geometry will be
exported as indexed, allowing vertices to be
shared. Non-indexed meshes are faster on
OpenGL 1.0.
Triangle strips
If checked, the exporter will convert
polygon meshes to triangle lists or triangle
strips if possible. This option is faster on all
platforms but should not be checked if you
are producing subdivision surfaces (see the
option Force subdivision surfaces). If not
checked, the Maya polygonal structure is
preserved in the OpenGL Performer
geometry. (Triangle stripping is not
performed on indexed meshes.)
Force subdivision surfaces
If checked, the exporter will convert
polygon meshes to subdivision surfaces in
the OpenGL Performer file. This works best
with low-count polygon meshes. Using
complex meshes can exhaust memory or
produce large models that are slow to
display.
Geometry Options
294
007-1680-100
The Maya Database Exporter
Table 7-14
Maya Export Options (continued)
Category
Texture Options
007-1680-100
Option
Description
Force polygon meshes
If checked, the exporter will convert
subdivision surfaces and NURBS models to
polygon meshes. The format of the output
meshes is controlled by the mesh export
options (see category Export geometry). If
not checked, subdivision surfaces and
NURBS from Maya are preserved in
OpenGL Performer.
Export textures
If checked, texturing and UV mapping is
enabled. Objects that are textured in Maya
will be output as textured in the OpenGL
Performer file. If you uncheck this box, the
exporter ignores all Maya texturing and
produces OpenGL Performer geometry
with materials only.
Make RGB texture files
If checked, all textures encountered in Maya
are converted to RGB format as part of the
export process. The Width and Height
sliders control the size of the RGB texture
images. This option should be checked if
you are using procedural textures in your
Maya scene or if you plan to use IRIX to
display your OpenGL Performer files.
295
7: Importing Databases
Exporting a Scene Using the Maya Embedded Language (MEL)
If you are familiar with MEL, the Maya scripting language, you can invoke the
OpenGL Performer export plug-in from within a script. This allows you to do batch
translation of Maya files to OpenGL Performer. The following sample MEL script
converts all of the Maya binary (.mb) files in a given directory to OpenGL Performer
binary (.pfb) format:
global proc int ExportAsPFB(string $indir, string $opts)
{
string $infile;
string $outfile;
string $files[];
string $s;
int $succeed = 1;
if ($opts == "")
$opts = "textures=1;width=256;height=256;indexed=1;tristrip=1;"
"gl1=1;preview=0;";
if ($indir == "")
$indir = "./";
setProject $indir;
$files=`getFileList -folder $indir -filespec "*.mb"`;
for ($infile in $files)
{
$outfile = $indir + substring($infile, 1, size($infile) - 3) +
".pfb";
$infile = $indir + $infile;
file -open -force $infile;
$s = `file -exportAll -type "PFBexport" -options $opts -force
$outfile`;
$succeed = `file -query -exists $s`;
if (!$succeed)
$s = "ERROR: failed to export " + $outfile + "\n";
else
$s = "exported " + $s + "\n";
print $s;
}
return $succeed;
}
296
007-1680-100
The Maya Database Exporter
Translation Details
This section describes the translation of Maya scenes to OpenGL Performer using the
following topics:
•
“Supported Maya Features”
•
“Hierarchy”
•
“Geometry”
•
“Shaders”
Supported Maya Features
Maya has many capabilities that cannot be represented in OpenGL Performer,
particularly in the area of shaders. Some of these features are simply ignored while others
may suffer quality loss. OpenGL Performer uses the OpenGL graphics pipeline, which
only allows light sources to illuminate objects directly. Light reflected from objects does
not illuminate other objects in the scene as it does with Maya. Consequently, high-quality
shading effects are often lost in the conversion.
Table 7-15 lists the Maya features that are supported by the exporter (column 1) and those
that are not supported (column 2).
Table 7-15
007-1680-100
Maya Features Supported by the Exporter
Supported Maya Features
Unsupported Maya Features
Polygonal meshes
Layered shaders
Subdivision surfaces
Ramp shaders
NURBS
Shading maps
Multitexturing
Anisotropic materials
Lambert, Blinn and Phong materials
Volumetric materials
Layered textures
Translucence
Procedural textures
Displacement mapping
Filtering, mipmapping
Quadratic, quartic, and Gaussian filtering
Reflection mapping
3D or cube map textures
297
7: Importing Databases
Table 7-15
Maya Features Supported by the Exporter (continued)
Supported Maya Features
Unsupported Maya Features
Cameras
Animation and skinning
Ambient, point, directional, and spot light sources
Area and volume light sources
Node instancing
Ray tracing
Shadow mapping
Custom shaders and mental ray
Bump mapping
Glow, fog, and motion blur
Vertex colors
Dynamics and fluids
Per-object light list
Switches, expressions, and scripting
Hierarchy
The exporter attempts to preserve the structure of the Maya scene graph as much as
possible in the OpenGL Performer output. Nodes that are instanced in Maya are also
shared in the OpenGL Performer hierarchy.
Maya and OpenGL Performer use different internal coordinate systems. This is not a
issue when importing a Maya scene into an OpenGL Performer application if you use the
whole scene because the root transformation corrects for the coordinate system
differences. However, if you take part of a scene from a .pfb file exported from Maya
and use it within another OpenGL Performer scene, the Maya objects may be oriented
incorrectly.
The following subsections describe translation details surrounding the scene graph
hierarchy:
298
•
Naming
•
Cameras
•
Light Sources
•
Sprites and Billboards
007-1680-100
The Maya Database Exporter
Naming
Node names are preserved so that Maya objects can be found by name in an OpenGL
Performer application. The name of an OpenGL Performer scene graph node is
composed of the file base name and the Maya object name. This is to permit multiple
Maya files to be imported without naming conflicts. For example, the Maya object
pCubeShape exported to file /usr/people/nola/myscene.pfb would be called
myscene.pCubeShape in the OpenGL Performer scene graph.
Cameras
Maya cameras are kept in the scene graph and any camera node may be used to view the
scene. In OpenGL Performer, the camera functionality is provided by the channel, which
is the root of the scene hierarchy. Maya cameras are exported as OpenGL Performer DCS
(dynamic coordinate system) nodes and keep their relative position within the hierarchy
so that they are accessible in OpenGL Performer applications. The root of the exported
scene graph includes the appropriate viewing transformation to display the scene from
the viewpoint of the current Maya viewport camera. The Maya background color
associated with the camera is not preserved in the OpenGL Performer scene.
Light Sources
OpenGL Performer has equivalents for ambient, directional, point, and spot light sources
because they are supported by the OpenGL pipeline. Area and volume lights are ignored
by the exporter. The exporter honors Maya per-object lighting. If you specify that only
certain lights should affect an object in Maya, this light list is associated with the OpenGL
Performer object. Because adding a light source in a real-time system can be
computationally costly, it is a good idea to use as few lights as possible when exporting
to OpenGL Performer.
Sprites and Billboards
A Maya sprite is an animated particle type that maps an image sequence onto a flat
rectangle. Sprites always face the camera and maintain their size when moved in the Z
direction. The sprite texture can be a single image or a sequence of images (texture
animation). Sprites are supported by the exporter.
Maya does not have an equivalent to the OpenGL Performer billboard that makes
arbitrary geometry face the camera at run time. Billboards are exported as DCS nodes.
007-1680-100
299
7: Importing Databases
Geometry
Maya has three ways to construct and represent geometry:
•
Indexed triangle meshes
•
Subdivision surfaces
•
NURBS
OpenGL Performer supports all three of these surface description methods directly. So,
Maya geometry can be displayed without a loss of fidelity.
The following subsections further describe geometry translation:
•
Polygonal Meshes
•
Subdivision Surfaces
•
UV Mapping
Polygonal Meshes
Polygonal meshes in Maya are indexed and share vertex, normal, and texture coordinate
data whenever possible. In Maya, different faces in the mesh can use different shaders.
Therefore, a single mesh can be associated with multiple appearances. In this case, one
mesh in Maya can produce several OpenGL Performer meshes, one for each appearance.
This is true of surfaces as well.
There are several ways to export Maya meshes to OpenGL Performer depending on how
you plan to use the geometry and the platform. Static geometry that is not modified at
run time can be highly optimized to display quickly. Meshes that will be dynamically
modified preserve the original Maya structure to make vertex manipulation at run time
easier and more efficient.
All exported meshes have locations and either normals or vertex colors depending on
how they were constructed in Maya. Texture coordinate sets are only included if texture
output is enabled and the mesh is textured. Per-vertex colors are not supported by the
exporter and will be ignored.
Subdivision Surfaces
Both Maya and OpenGL Performer subdivision surfaces use the Catmull-Clark
algorithm. Therefore, run-time tesselation should produce results similar to Maya.
300
007-1680-100
The Maya Database Exporter
OpenGL Performer subdivision surfaces allow multiple materials on a single mesh and
will preserve connectivity.
Although Maya NURBS shapes can use multiple shaders, the OpenGL Performer
NURBS geode only supports one material. The exporter does not support the use of
multiple shaders across a NURBS surface.
You can export a surface as a polygonal mesh and the exporter will use Maya to tesselate
the surface.
UV Mapping
The UV mapping process transforms or generates texture coordinates for a mesh. Static
UV mapping is applied to the Maya texture coordinates at export time. Environment
mapping dynamically generates texture coordinates at run time based on the location of
the viewer. Texture coordinate animation is not supported by the exporter.
Table 7-16 describes how the exporter supports the various UV mapping methods.
Table 7-16
Maya Exporter Support for UV Mapping Methods
UV Map Method
Exporter Support
place2dTexture
2D translation, scaling and rotation of the texture coordinates done at export
time.
envSphere
Spherical environment mapping. Texture coordinates are generated by
OpenGL Performer at run time.
envCube
Cubic environment mapping. Texture coordinates are generated by
OpenGL Performer at run time.
Place3dTexture
3D matrix applied to texture coordinates at run time.
envSky
Ignored.
envChrome
Ignored.
envBall
Ignored.
Shaders
The following subsections describe how the exporter supports Maya shaders:
007-1680-100
301
7: Importing Databases
•
Materials
•
Textures
•
Texture Effects
•
Procedural Textures
•
Layered Textures
Materials
Lambert, Blinn, and Phong materials are supported by the exporter but not all of the
material properties are honored, as shown in Table 7-17.
Table 7-17
302
Maya Exporter Support for Material Properties
Material Property
Exporter Support
Color
Becomes diffuse material color.
Transparency
Becomes material transparency. You cannot specify a map here to be used
as an alpha channel. The texture alpha channel must be included in the
texture file.
Ambient color
Becomes ambient material color. If a texture map is used here, it is added
to the overall color and light sources do not affect it.
Incandescence
Becomes emissive material color. If a texture map is used here, it is added
to the overall color.
Diffuse
Is used as a multiplier for diffuse material color.
Translucence
Ignored.
Glow intensity
Ignored.
Hide source
Ignored.
Matte opacity
Ignored.
Ray trace options
Ignored.
007-1680-100
The Maya Database Exporter
Textures
All textures used with OpenGL Performer must come from bitmap files. Because
OpenGL Performer does not support all texture formats on every platform, use the .rgb
format for textures if you plan to use your exported files on multiple platforms.
OpenGL Performer multitexturing support is hardware-based and varies across
platforms. On some older IRIX systems, multitexturing is not supported. Newer
hardware can have two or four texture units. The Maya exporter supports a maximum
of eight textures per object. It is possible to produce .pfb files that will not look the same
on all platforms. If you use more than eight textures on a single mesh, the exporter will
ignore the extra textures and produce an error message in the log file.
Texture Effects
Maya allows you to correctly color your textures by adding and/or multiplying the
texture by a color before applying it (using the color gain and color offset
attributes). You can also invert or filter the texels. These texture effects are only preserved
if you convert your textures to RGB. The exporter captures the texture after these
computations have been performed.
Table 7-18 describes how the exporter supports the various the Maya texture properties.
Table 7-18
007-1680-100
Maya Exporter Support for Texture Properties
Texture Property
Exporter Support
Image name
Name of bitmap file containing texture data.
Color gain
Ignored unless the Make RGB option is enabled.
Color offset
Ignored unless the Make RGB option is enabled.
Invert
Ignored the unless Make RGB option is enabled.
Wrap
Becomes texture wrap mode.
Filter
Only mipmap and box filters may be used.
Hardware cycling
If animation export is enabled, a texture animation is produced.
Otherwise, it is ignored.
Filter offset
Ignored.
Blend
Ignored.
303
7: Importing Databases
Table 7-18
Maya Exporter Support for Texture Properties (continued)
Texture Property
Exporter Support
Alpha gain
Ignored.
Alpha offset
Ignored.
Alpha is luminance
Ignored.
Default color
Ignored.
Color remap
Ignored.
Procedural Textures
Maya has a rich library of procedural textures that can dynamically create texel patterns
without requiring bitmap files. When the exporter encounters a procedural texture, it
produces a reference to a bitmap file containing a snapshot of the procedural texture from
Maya. If the Make RGB option is enabled, the exporter creates a separate RGB bitmap
file for every texture used in the scene. This step only needs to be done when you add
new textures or change the construction options on an existing procedural texture.The
name of the RGB file is constructed by using the name of the Maya texture object with a
.rgb suffix. Consequently, if you export multiple files that use the same object names,
the exporter will use the same filename for multiple textures. Consequently, the same
filenames could cause problems if the textures are not the same.
Procedural textures are emulated in OpenGL Performer using 2D UV coordinates and
the texture matrix. It is not always possible to get the same results as Maya, particularly
with 3D textures. The exporter uses the first UV set on the mesh to map the snapshot of
the procedural texture. By manipulating these UVs, you can often improve the real-time
appearance of procedural textures.
Layered Textures
Layered textures in Maya allow the artist to blend several different textures together to
produce an output color. The exporter supports layered textures but OpenGL Performer
cannot implement all of the blend modes. Only the None, Over, In, Add and Multiply
blend modes are honored. All other modes default to Multiply.
304
007-1680-100
Chapter 8
8. Geometry
All libpr geometry is defined by modular units of prmitives that employ a flexible
specification method. The following two classes allow you to group these basic
primitives:
•
“pfGeoSet (Geometry Set)” on page 305
•
“pfGeoArray (Geometry Array)” on page 319
This chapter describes these two classes and the following two topics:
•
“Optimizing Geometry for Rendering” on page 329
•
“Rendering 3D Text” on page 335
pfGeoSet (Geometry Set)
A pfGeoSet is a collection of geometry that shares certain characteristics. All items in a
pfGeoSet must be of the same primitive type (whether they are points, lines, or triangles)
and share the same set of attribute bindings (you cannot specify colors-per-vertex for
some items and colors-per-primitive for others in the same pfGeoSet). A pfGeoSet forms
primitives out of lists of attributes that may be either indexed or nonindexed. An indexed
pfGeoSet uses a list of unsigned short integers to index an attribute list. (See “Attributes”
on page 313 for information about attributes and bindings.)
Indexing provides a more general mechanism for specifying geometry than hard-wired
attribute lists and also has the potential for substantial memory savings as a result of
shared attributes. Nonindexed pfGeoSets are sometimes easier to construct, usually a bit
faster to render, and may save memory (since no extra space is needed for index lists) in
situations where vertex sharing is not possible. A pfGeoSet must be either completely
indexed or completely nonindexed; it is not valid to have some attributes indexed and
others nonindexed.
007-1680-100
305
8: Geometry
Note: libpf applications can include pfGeoSets in the scene graph with the pfGeode
(Geometry Node).
Table 8-1 lists a subset of the routines that manipulate pfGeoSets.
Table 8-1
306
pfGeoSet Routines
Routine
Description
pfNewGSet()
Create a new pfGeoSet.
pfDelete()
Delete a pfGeoSet.
pfCopy()
Copy a pfGeoSet.
pfGSetGState()
Specify the pfGeoState to be used.
pfGSetGStateIndex()
Specify the pfGeoState index to be used.
pfGSetNumPrims()
Specify the number of primitive items.
pfGSetPrimType()
Specify the type of primitive.
pfGSetPrimLengths()
Set the lengths array for strip primitives.
pfGetGSetPrimLength()
Get the length for the specified strip primitive.
pfGSetAttr()
Set the attribute bindings.
pfGSetMultiAttr()
Set multi-value attributes (for example, multi-texture coordinates).
pfGSetDrawMode()
Specify draw mode (for example, flat shading or wireframe).
pfGSetLineWidth()
Set the line width for line primitives.
pfGSetPntSize()
Set the point size for point primitives.
pfGSetHlight()
Specify highlighting type for drawing.
pfDrawGSet()
Draw a pfGeoSet.
pfGSetBBox()
Specify a bounding box for the geometry.
pfGSetIsectMask()
Specify an intersection mask for pfGSetIsectSegs().
pfGSetIsectSegs()
Intersect line segments with pfGeoSet geometry.
007-1680-100
pfGeoSet (Geometry Set)
Table 8-1
pfGeoSet Routines (continued)
Routine
Description
pfQueryGSet()
Determine the number of triangles or vertices.
pfPrint()
Print the pfGeoSet contents.
Primitive Types
All primitives within a given pfGeoSet must be of the same type. To set the type of all
primitives in a pfGeoSet named gset, call pfGSetPrimType(gset, type). Table 8-2 lists the
primitive type tokens, the primitive types that they represent, and the number of vertices
in a coordinate list for that type of primitive.
Table 8-2
Geometry Primitives
Token
Primitive Type
Number of Vertices
PFGS_POINTS
Points
numPrims
PFGS_LINES
Independent line segments
2 * numPrims
PFGS_LINESTRIPS
Strips of connected lines
Sum of lengths array
PFGS_FLAT_LINESTRIPS
Strips of flat-shaded lines
Sum of lengths array
PFGS_TRIS
Independent triangles
3 * numPrims
PFGS_TRISTRIPS
Strips of connected triangles
Sum of lengths array
PFGS_FLAT_TRISTRIPS
Strips of flat-shaded triangles
Sum of lengths array
PFGS_TRIFANS
Fan of conected triangles
Sum of lengths array
PFGS_FLAT_TRIFANS
Fan of flat-shaded triangles
Sum of lengths array
PFGS_QUADS
Independent quadrilaterals
4 * numPrims
PFGS_POLYS
Independent polygons
Sum of lengths array
The parameters in the last column denote the following:
numPrims
007-1680-100
The number of primitive items in the pfGeoSet, as set by
pfGSetNumPrims().
307
8: Geometry
lengths
The array of strip lengths in the pfGeoSet, as set by
pfGSetPrimLengths() (note that length is measured here in terms of
number of vertices).
Connected primitive types (line strips, triangle strips, and polygons) require a separate
array that specifies the number of vertices in each primitive. Length is defined as the
number of vertices in a strip for STRIP primitives and is the number of vertices in a
polygon for the POLYS primitive type. The number of line segments in a line strip is
numVerts – 1, while the number of triangles in a triangle strip and polygon is numVerts –
2. Use pfGSetPrimLengths() to set the length array for strip primitives.
The number of primitives in a pfGeoSet is specified by pfGSetNumPrims(gset, num). For
strip and polygon primitives, num is the number of strips or polygons in gset.
pfGeoSet Draw Mode
In addition to the primitive type, pfGSetDrawMode() further defines how a primitive is
drawn. Triangles, triangle strips, quadrilaterals, and polygons can be specified as either
filled or as wireframe, where only the outline of the primitive is drawn. Use the
PFGS_WIREFRAME argument to enable or disable wireframe mode. Another argument,
PFGS_FLATSHADE, specifies that primitives should be shaded. If flat shading is
enabled, each primitive or element in a strip is shaded with a single color.
PFGS_COMPILE_GL
At the next draw for each pfState, compile gset’s geometry into a GL
display list and subsequently render the display list.
PFGS_DRAW_GLOBJ
Select the rendering of an already created display list but do not force a
recompile.
PFGS_PACKED_ATTRS
Use the gset’s packed attribute arrays, set with the
PFGS_PACKED_ATTRS to pfGSetAttr, to render geometry with GL
vertex arrays.
The pfGeoSets are normally processed in immediate mode, which means that
pfDrawGSet() sends attributes from the user-supplied attribute arrays to the Graphics
Pipeline for rendering. However, this kind of processing is subject to some overhead,
particularly if the pfGeoSet contains few primitives. In some cases it may help to use GL
display lists (this is different from the libpr display list type pfDispList) or compiled
308
007-1680-100
pfGeoSet (Geometry Set)
mode. In compiled mode, pfGeoSet attributes are copied from the attribute lists into a
special data structure called a display list during a compilation stage. This data structure
is highly optimized for efficient transfer to the graphics hardware. However, compiled
mode has some major disadvantages:
•
Compilation is usually costly.
•
A GL display list must be recompiled whenever its pfGeoSet’s attributes change.
•
The GL display list uses a significant amount of extra host memory.
In general, immediate mode will offer excellent performance with minimal memory
usage and no restrictions on attribute volatility, which is a key aspect in may advanced
applications. Despite this, experimentation may show databases or machines where
compiled mode offers a performance benefit.
To enable or disable compiled mode, call pfGSetDrawMode() with the
PFGS_COMPILE_GL token. When enabled, compilation is delayed until the next time
the pfGeoSet is drawn with pfDrawGSet(). Subsequent calls to pfDrawGSet() will then
send the compiled pfGeoSet to the graphics hardware.
To select a display list to render, without recompiling it, use pfGSetDrawMode() with
the token PFGS_DRAW_GLOBJ.
Packed Attributes
Packed attributes is an optimized way of sending formatted data to the graphics pipeline
under OpenGL that does not incur the same memory overead or recompilation burden
as GL display lists. To render geometry with packed attributes, use the
pfGSetDrawMode(PFGS_PACKED_ATTRS) method when using OpenGL. This
pfGSetAttr list includes the currently bound PER_VERTEX vertex attribute data packed
into a single nonindexed array. When specifying a packed attribute array, the optional
vertex attributes, colors, normals, and texture coordinates, can be NULL. This array, like
the other attribute arrays, is then shared betweenOpenGL Performer, the GL, and
accessible by the user. Optionally, you can put your vertex coordinates in this packed
array but in this case the vertices must be duplicated in the normal coordinate array
because vertex coordinate data is used internally for other nondrawing operations such
as intersections and computation of bounding geometry. Packed attribute arrays also
allow OpenGL Performer to extend the vertex attribute types accepted by pfGeoSets.
There are several base formats that expect all currently bound attributes of specified data
type (unsigned byte, short, or float) to be in the attribute array. Attributes specified by the
format but not bound to vertices are assumed to not be present and the present data is
007-1680-100
309
8: Geometry
packed with the data for each vertex starting on a 32-bit word-aligned boundary. Then,
there are several derived formats that let you put some attribute data in the packed array
while leaving the rest in the normal individual coordinate attribute arrays. Table 8-3
shows the different base formats supported.
Table 8-3
pfGeoSet PACKED_ATTR Formats
Format
Description
PFGS_PA_C4UBN3ST2FV3F
Accepts all currently bound coordinate attributes; colors are
unsigned bytes; normals are shorts. Vertices are duplicated in
the packed attribute array.
PFGS_PA_C4UBN3ST2F
Vertices are in the normal coordinate array.
PFGS_PA_C4UBT2F
Normals and vertices are in the normal coordinate array.
PFGS_PA_C4UBN3ST2SV3F
All bound coordinate attributes are in the packed attribute
array. Colors are unsigned bytes, normals are shorts, and
texture coordinates are unsigned shorts.
PFGS_PA_C4UBN3ST3FV3F
Texture coordinates are 3D floats.
PFGS_PA_C4UBN3ST3SV3F
Texture coordinates are 2D shorts.
To create packed attributes, you can use the utility pfuTravCreatePackedAttrs(), which
traverses a scene graph to create packed attributes for pfGeoSets and, optionally,
pfDelete redundant attribute arrays. This utility packs the pfGeoSet attributes using
pfuFillGSetPackedAttrs(). Examples of packed attribute usage can be seen in
/usr/share/Performer/src/pguide/libpr/C/packedattrs.c and
/usr/share/Performer/src/sample/C/perfly.c and
/usr/share/Performer/src/sample/C++/perfly.C for IRIX and Linux and in
%PFROOT%\Src\pguide\libpr\C\packedattrs.c,
%PFROOT%\Src\sample\C\perfly.c, and
%PFROOT%\Src\sample\C++\perfly.C for Microsoft Windows.
Primitive Connectivity
A pfGeoSet requires a coordinate array that specifies the world coordinate positions of
primitive vertices. This array is either indexed or not, depending on whether a
coordinate index list is supplied. If the index list is supplied, it is used to index the
coordinate array; if not, the coordinate array is interpreted in a sequential order.
310
007-1680-100
pfGeoSet (Geometry Set)
A pfGeoSet’s primitive type dictates the connectivity from vertex to vertex to define
geometry. Figure 8-1 shows a coordinate array consisting of four coordinates, A, B, C,
and D, and the geometry resulting from different primitive types. This example uses
index lists that index the coordinate array.
007-1680-100
311
8: Geometry
Note: Flat-shaded line strip and flat-shaded triangle strip primitives have the vertices
listed in the same order as for the smooth-shaded varieties.
O
1
2
3
...
n
D
C
A
B
Primitive
type
Points
Vertex list
XA, YA, ZA
XB, YB, ZB
XC, YC, ZC
XD, YD, ZD
XN, YN, ZN
Line strips
Line segments
Geometry
Index list
Primitive
type
0
0
0
2
3
1
3
1
3
2
2
1
3
0
1
3
2
2
1
0
Independent Quadrilaterals
triangles
Triangle strips
Polygons
Geometry
Index list
Figure 8-1
312
0
0
0
0
1
1
1
1
3
2
3
2
3
3
2
3
1
...
...
2
n
n
Primitives and Connectivity
007-1680-100
pfGeoSet (Geometry Set)
Attributes
The definition of a primitive is not complete without attributes. In addition to a primitive
type and count, a pfGeoSet references four attribute arrays (see Figure 8-2):
•
Colors (red, green, blue, alpha)
•
Normals (Nx, Ny, Nz)
•
Texture coordinates (S, T)—multiple arrays for multitexture.
•
Vertex coordinates (X, Y, Z)
(A pfGeoState is also associated with each pfGeoSet; see Chapter 12, “Graphics State” for
details.) The four components listed above can be specified with pfGSetAttr().
Multivalue attributes (texture coordinates) can be specified using pfGSetMultiAttr() or
pfGSetAttr(). Using zero as the index parameter for pfGSetMultiAttr() is equivalent to
calling pfGSetAttr(). Attributes may be set in two ways: by indexed specification—using
a pointer to an array of components and a pointer to an array of indices; or by direct
specification—providing a NULL pointer for the indices, which indicates that the indices
are sequential from the initial value of zero. The choice of indexed or direct components
applies to an entire pfGeoSet; that is, all of the supplied components within one pfGeoSet
must use the same method. However, you can emulate partially indexed pfGeoSets by
using indexed specification and making each nonindexed attribute’s index list be a singly
shared “identity mapping” index array whose elements are 0, 1, 2, 3,…, N–1, where N is
the largest number of attributes in any referencing pfGeoSet. (You can share the same
array for all such emulated pfGeoSets.) The direct method avoids one level of indirection
and may have a performance advantage compared with indexed specification for some
combinations of CPUs and graphics subsystems.
Note: Use pfMalloc() to allocate your arrays of attribute data. This allows OpenGL
Performer to reference-count the arrays and delete them when appropriate. It also allows
you to easily put your attribute data into shared memory for multiprocessing by
specifying an arena such as pfGetSharedArena() to pfMalloc(). While perhaps
convenient, it is very dangerous to specify pointers to static data for pfGeoSet attributes.
Early versions of OpenGL Performer permitted this but it is strongly discouraged and
may have undefined and unfortunate consequences.
Attribute arrays can be created through pfFlux to support the multiprocessed generation
of the vertex data for a dynamic object, such as ocean waves, or morphing geometry.
pfFlux will automatically keep separate copies of data for separate proceses so that one
007-1680-100
313
8: Geometry
process can generate data while another draws it. The pfFluxed buffer can be handed
directly to pfGSetAttr() or pfGSetMultiAttr(). In fact, the entire pfGeoSet can be
contained in a pfFlux. Index lists cannot be pfFluxed. See Chapter 19, “Dynamic Data,”
for more information on pfFlux.
0: R
G
1: R B A
G
2: R B A
GB
A
lor
Co ay
r
ar
Ge
o,1
,3,2
,8,2
lor
Co ex
ind
...
rd
coo
ord
T e x r r a y e x c oy
ord
a T r r a e x c oy
d
0: S
a
T rra
oor
x ca y
a
T 0: S
e
T
0: S
1: S
T
arr
T
T 1:
2: S
T
ST
2: S
T
1: S
T
2: S
T
0: S
T
1: S
T
2: S
T
dxc
oor
314
p
et
0:n
xn
1:n y nz
x
2:n ny nz
xn
yn
z
3,1
,8,3
,2..
0: X
Y
1: X Z
Y
2: X Z
YZ
11,
4,8
,2,6
l
a
rm
Nodex
in
l
.
d
0,1 d c o o rT e n d e x
,r2 x
i
0,1 d x c o o T,3e,4n,5d.e x
,r2T,3e e x i ..
o
0,1
o
,2,3e x c x ,i4n,d5...
T ,4n,d5 e
0,1
,2,3
i ...
,4,5
...
Figure 8-2
oSt
at
prim e
itiv
e ty
pe
prim
itiv
ec
oun
t
col
or a
rra
y
col
or i
nde
x
nor
ma
l ar
ray
nor
ma
l in
dex
tex
tur
ec
oor
da
tex
rra
tur
y
ec
oor
d in
ver
dex
tex
coo
rd a
ver
rra
tex
y
coo
rd i
nde
x
oS
fGe
a
rm
No ray
ar
x
rte
Ve ray
ar
x
rte
Ve ex
ind
...
pfGeoSet Structure
007-1680-100
pfGeoSet (Geometry Set)
Note: When using multiple texture-coordinate arrays, pfGeoSet recognizes
texture-coordinate arrays starting at the first array (index of 0) and ending immediately
before the first index with a NULL array. In other words, specifying texture-coordinate
arrays using pfGSetMultiAttr() for indices 0, 1, and 3 is equivalent to specifying
texture-coordinate arrays for only indices 0 and 1. When using pfTexGen to
automatically generate texture coordinates for some texture units, the application should
not interleave texture units with texture coordinates and texture units with pfTexGen.
Texture units with texture coordinates should come before texture units with pfTexGen.
This is an implementation limitation and may be removed in future releases.
Attribute Bindings
Attribute bindings specify where in the definition of a primitive an attribute has effect.
You can leave a given attribute unspecified; otherwise, its binding location is one of the
following:
•
Overall (one value for the entire pfGeoSet)
•
Per primitive
•
Per vertex
Only certain binding types are supported for some attribute types.
Table 8-4 shows the attribute bindings that are valid for each type of attribute.
Table 8-4
Attribute Bindings
Binding Token
Color
Normal
Texture Coordinate
Coordinate
PFGS_OVERALL
Yes
Yes
No
No
PFGS_PER_PRIM
Yes
Yes
No
No
PFGS_PER_VERTEX
Yes
Yes
Yes
Yes
PFGS_OFF
Yes
Yes
Yes
No
Attribute lists, index lists, and binding types are all set by pfGSetAttr().
007-1680-100
315
8: Geometry
For FLAT primitives (PFGS_FLAT_TRISTRIPS,PFGS_FLAT_TRIFANS,
PFGS_FLAT_LINESTRIPS), the PFGS_PER_VERTEX binding for normals and colors has
slightly different meaning. In these cases, per-vertex colors and normals should not be
specified for the first vertex in each line strip or for the first two vertices in each triangle
strip since FLAT primitives use the last vertex of each line segment or triangle to compute
shading.
Indexed Arrays
A cube has six sides; together those sides have 24 vertices. In a vertex array, you could
specify the primitives in the cube using 24 vertices. However, most of those vertices
overlap. If more than one primitive can refer to the same vertex, the number of vertices
can be streamlined to 8. The way to get more than one primitive to refer to the same
vertex is to use an index; three vertices of three primitives use the same index which
points to the same vertex information. Adding the index array adds an extra step in the
determination of the attribute, as shown in Figure 8-3.
316
007-1680-100
pfGeoSet (Geometry Set)
pfGeoSet
StripLengths
PrimCoords
ColorBind
NormalBind
TexCoordBind
n1
n2
n3
.
.
.
CoordSet
ColorSet
NormalSet
TexCoordSet
CoordIndexSet
ColorIndexSet
NormalIndexSet
TextCoordIndexSet
< x, y, z >
.
.
.
n1
n2
n3
.
.
.
Figure 8-3
< r, g, b >
.
.
.
n1
n2
n3
.
.
.
< nx, ny, nz >
.
.
.
n1
n2
n3
.
.
.
< x, y, z >
.
.
.
n1
n2
n3
.
.
.
Indexing Arrays
Indexing can save system memory, but rendering performance is often lost.
When to Index Attributes
The choice of using indexed or sequential attributes applies to all of the primitives in a
pfGeoSet; that is, all of the primitives within one pfGeoSet must be referenced
sequentially or by index; you cannot mix the two.
The governing principle for whether to index attributes is how many vertices in a
geometry are shared. Consider the following two examples in Figure 8-4, where each dot
marks a vertex.
007-1680-100
317
8: Geometry
Figure 8-4
Deciding Whether to Index Attributes
In the triangle strip, each vertex is shared by two adjoining triangles. In the square, the
same vertex is shared by eight triangles. Consider the task that is required to move these
vertices when, for example, morphing the object. If the vertices were not indexed, in the
square, the application would have to look up and alter eight triangles to change one
vertex.
In the case of the square, it is much more efficient to index the attributes. On the other
hand, if the attributes in the triangle strip were indexed, since each vertex is shared by
only two triangles, the index look-up time would exceed the time it would take to simply
update the vertices sequentially. In the case of the triangle strip, rendering is improved
by handling the attributes sequentially.
The deciding factor governing whether to index attributes relates to the number of
primitives that share the same attribute: if attributes are shared by many primitives, the
attributes should be indexed; if attributes are not shared by many primitives, the
attributes should be handled sequentially.
pfGeoSet Operations
There are many operations you can perform on pfGeoSets. pfDrawGSet() “draws “ the
indicated pfGeoSet by sending commands and data to the Geometry Pipeline, unless
OpenGL Performer’s display-list mode is in effect. In display-list mode, rather than
sending the data to the pipeline, the current pfDispList “captures” the pfDrawGSet()
command. The given pfGeoSet is then drawn along with the rest of the pfDispList with
the pfDrawDList() command.
When the PFGS_COMPILE_GL mode of a pfGeoSet is not active (pfGSetDrawMode()),
pfDrawGSet() uses rendering loops tuned for each primitive type and attribute binding
combination to reduce CPU overhead in transferring the geometry data to the hardware
pipeline. Otherwise, pfDrawGSet() sends a special, compiled data structure.
318
007-1680-100
pfGeoArray (Geometry Array)
Table 8-1 on page 306 lists other operations that you can perform on pfGeoSets. pfCopy()
does a shallow copy, copying the source pfGeoSet’s attribute arrays by reference and
incrementing their reference counts. pfDelete() frees the memory of a pfGeoSet and its
attribute arrays (if those arrays were allocated with pfMalloc() and provided their
reference counts reach zero). pfPrint() is strictly a debugging utility and will print a
pfGeoSet’s contents to a specified destination. pfGSetIsectSegs() allows intersection
testing of line segments against the geometry in a pfGeoSet; see “Intersecting with
pfGeoSets” in Chapter 22 for more information on that function.
pfGeoArray (Geometry Array)
The pfGeoArray is a new OpenGL Performer data structure aimed at replacing the class
pfGeoSet. Conceptually, pfGeoArrays are very similar to pfGeoSets, but they allow you
to define new sets of attributes in addition to the standard vertex coordinates, normals,
texture coordinates, and colors. These attributes can be used by vertex or fragment
programs applied to the primitives (see “Using OpenGL Performer with GPUs” on
page 532). Also, pfGeoArrays are rendered using vertex arrays and vertex objects,
making the rendering much more efficient. pfGeoArrays can be up to 10 times faster than
pfGeoSets on Onyx4 or Prism systems.
Each pfGeoArray is a collection of geometry with one primitive type, such as points,
lines, or triangles. Vertex coordinates, normals, colors, texture coordinates, and
user-defined attributes are used to specify the primitives. There are two ways to specify
the attributes. First, each attribute is specified per vertex, there is no concept of an
attribute per primitive or an overall attribute. Second, you can use a single list of
unsigned integers to index all attributes of a pfGeoArray.
Indexing provides a more general mechanism for specifying geometry than hardwired
attribute lists and, in many cases, provides substantial memory savings due to shared
attributes. Nonindexed pfGeoArrays are sometimes easier to construct and may exhibit
better caching behavior. Indexing is often a desirable approach especially when your
primitives are sharing many attributes (such as having the same normal for each face).
Also, if you have a primitive with many triangle strips, it is better to create a single
pfGeoArray containing indexed triangles than to have a set of short pfGeoArrays, each
containing one triangle strip.
007-1680-100
319
8: Geometry
This section contains the following topics:
•
“Creating pfGeoArrays” on page 320
•
“pfGeoArray Attributes” on page 320
•
“pfGeoArray Attribute Types” on page 323
•
“pfGeoArray Primitive Types” on page 323
•
“Example Code” on page 324
•
“Converting pfGeoSets to pfGeoArrays” on page 328
Creating pfGeoArrays
The function pfNewGArray() creates and returns a handle to a pfGeoArray. The
parameter arena specifies a malloc() arena out of which the pfGeoArray is allocated or
NULL for allocation off the process heap. pfGeoArrays can be deleted with pfDelete().
The call new(arena) allocates a pfGeoArray from the specified memory arena, or from the
process heap if arena is NULL. The new() call allocates a pfGeoArray from the default
memory arena (see the man page for pfGetSharedArena). Like other pfObjects,
pfGeoArrays cannot be automatically created statically on the stack or in arrays. Delete
pfGeoArrays with pfDelete() rather than with the delete operator.
pfGeoArray Attributes
The function pfGArrayAddAttr() adds a new attribute to the list of attributes of a
pfGeoArray. This list is initially empty. An attribute is specified by its attribute type and
the following parameters:
320
size
Specifies the number of coordinates per vertex; It must be 2, 3, or 4.
type
Specifies the type of each component in the attribute data. It is one of
GL_DOUBLE, GL_FLOAT, GL_INT, GL_UNSIGNED_INT,
GL_UNSIGNED_SHORT, GL_SHORT, GL_UNSIGNED_BYTE, and
GL_BYTE.
stride
Specifies the byte offset between consecutive vertex data. It is usually 0.
pointer
Specifies a pointer to the attribute data.
007-1680-100
pfGeoArray (Geometry Array)
You can modify the name, size, the data type, the stride, and the data pointer for any
existing attribute using the following functions:
•
pfVAttrName()
•
pfVAttrPtr()
•
pfVAttrStride()
•
pfVAttrDataType()
•
pfVAttrSize()
You can also remove an attribute using the function pfGArrayRemoveAttr().
Multitexturing is supported by adding multiple PFGA_TEX_ARRAY vertex attributes and
specifying different stages, as shown in the following example:
pfGeoArray *gArray = pfNewGArray();
pfGArrayMultiAttr(gArray, PFGA_TEX_ARRAY, 0, 2, GL_FLOAT, 0,
baseCoords);
pfGArrayMultiAttr(gArray, PFGA_TEX_ARRAY, 1, 2, GL_FLOAT, 0,
bumpCoords);
/* set name for the two sets of tex coords just assigned */
pfVArrayName( pfGArrayQueryAttrTypeStage(gArray, PFGA_TEX_ARRAY, 0),
"base texture coords");
pfVArrayName( pfGArrayQueryAttrTypeStage(gArray, PFGA_TEX_ARRAY, 1),
"bump texture coords");
Note that since the pfGeoArray attributes are rendered in the order they were added, it
is possible to interleave the attributes with your own callbacks. To do so, create a special
"callback" type with a function mask 0 (no callback data) or a function mask 0x1
(callback data is used).
It is possible to index the attributes, although (in contrast to pfGeoSets) a single index list
is used for all attributes. The optional attribute index list is a list of unsigned short
integers. The index list is specified using the function pfGArrayIndexArray().
If attribute and index lists are allocated from the pfMalloc routines, pfGArrayAddAttr()
and pfGArrayAttrPtr() will correctly update the reference counts of the lists. Specifically,
they will decrement the reference counts of the old lists and increment the reference
counts of the new lists. It will not free any lists whose reference counts reach 0. When a
007-1680-100
321
8: Geometry
pfGeoArray is deleted with pfDelete(), all pfMalloc'ed lists will have their reference
counts decremented by one and will be freed if their count reaches 0.
When pfGeoArrays are copied with pfCopy(), all pfMalloc'ed lists of the source
pfGeoArray will have their reference counts incremented by one and those pfMalloc'ed
lists of the destination pfGeoArray will have their reference counts decremented by one.
The function pfCopy() copies lists only by reference (only the pointer is copied) and will
not free any lists whose reference counts reach 0.
Attribute data may be any of the following types of memory:
•
Data allocated with pfMalloc
This is the usual and recommended memory type for pfGeoArray index and
attribute arrays.
•
Static, malloc(), amalloc(), usmalloc(), and similar data (non-pfMalloc'ed data)
This type of memory is not generally recommended since it does not support
reference counting or other features provided by pfMalloc. In particular, do not use
static data because it may result in segmentation violations.
•
pfFlux memory
In a pipelined, multiprocessing environment, a pfFlux provides multiple data
buffers, which allow frame-accurate data modifications to pfGeoArray attribute
arrays like coordinates (facial animation) and texture coordinates (ocean waves,
surf). The functions pfGArrayAddAttr() and pfGArrayAttrPtr() will accept a
pfFlux* or pfFluxMemory* for the attribute list (index lists do not support pfFlux)
and the pfGeoArray will select the appropriate buffer when rendered or intersected.
See the man page for pfFlux for more details.
Since pfGeoArrays are cached using vertex array objects, if you want to animate
some attributes, you need to either disable caching using the function
pfGArrayAllowCache() or call the function pfGArrayUpdateData() each time you
change any of the attribute data.
•
pfCycleBuffer and pfCycleMemory
Note that pfCycleBuffer has been obsoleted by pfFlux. See the man page for
pfCycleBuffer for more details.
OpenGL Performer allows mixing pfMalloc'ed, pfFlux, and pfCycleBuffer attributes on
a single pfGeoArray.
322
007-1680-100
pfGeoArray (Geometry Array)
pfGeoArray Attribute Types
When a new pfGeoArray is created, it has no default attribute. When adding a new
attribute, you must specify the type of the attribute— that is, whether it specifies one of
the following:
•
A vertex coordinates (PFGA_COORD_ARRAY)
•
A normal vector (PFGA_NORMAL_ARRAY)
•
A color (PFGA_COLOR_ARRAY)
•
A texture coordinate (PFGA_TEX_ARRAY)
•
A generic user-defined attribute (PFGA_GENERIC_ARRAY)
Attribute types are identified by their type, their name (a string), and the associated
texture stage (if applicable). A new attribute type can be added using the function
pfGArrayAddAttrType(). The parameter type is one of tokens just cited. The name can
be any arbitrary string and if one is not set, then depending on the array type, a default
name will be used ("vertex", "normal", "color", "texture coord" or "generic").
The parameter stage defines the associated texture stage. In the case of attributes of
type PFGA_GENERIC_ARRAY, the attributes are applied using the function
glVertexAttribPointerARB().
The following example code adds two sets of texture coordinates and one set of vertices
to a pfGeoArray:
pfVertexAttr *vAttr, tAttrs[2];
pfGeoArray *gArray = ...;
vAttr
= pfGArrayAddAttrType(gArray, PFGA_COORD_ARRAY, "vertices", 0);
tAttr[0] = pfGArrayAddAttrType(gArray, PFGA_TEX_ARRAY, "texCoord0", 0);
tAttr[1] = pfGArrayAddAttrType(gArray, PFGA_TEX_ARRAY, "texCoord1", 1);
pfGeoArray Primitive Types
A primitive is a single point, line segment, line strip, triangle, triangle strip, quad, or
polygon depending on the primitive type. The primitive type dictates how the
coordinate and coordinate index lists are interpreted to form geometry. The function
pfGSetPrimType() specifies the type of primitives found in a pfGeoArray.
The following example shows how to set up a nonindexed, TRISTRIP pfGeoArray:
007-1680-100
323
8: Geometry
/* Set up a nonindexed, TRISTRIP pfGeoArray */
garray = pfNewGArray(NULL);
pfGSetPrimType(garray, PFGS_TRISTRIPS);
pfGSetNumPrims(garray, 1);
lengths[0] = 4;
pfGSetPrimLengths(gset, lengths);
coords = (pfVec3*) pfMalloc(sizeof(pfVec3) * 4, NULL);
colors = (pfVec4*) pfMalloc(sizeof(pfVec4) * 4, NULL);
pfGArraySetAttr(garray, PFGA_COORD_ARRAY, 3, GL_FLOAT, 0, coords);
pfGArraySetAttr(garray, PFGA_COLOR_ARRAY, 4, GL_FLOAT, 0, colors);
The function pfGetGSetClassType() returns the pfType* for the class
pfShaderProgram. The pfType* returned by pfGetGSetClassType() is the same as the
pfType* returned by invoking pfGetType(), the virtual function getType() on any
instance of class pfShaderProgram. Because OpenGL Performer allows subclassing of
built-in types when decisions are made based on the type of an object, use pfIsOfType()
the member function isOfType() to test if an object is of a type derived from an
OpenGL -Performer type rather than to test for strict equality of the pfType*s.
Example Code
The following example shows one way to create a pfGeoArray defining a hexahedron
(cube).
static pfVec3 coords[] =
{
{-1.0, -1.0, 1.0}, /* front */
{ 1.0, -1.0, 1.0},
{ 1.0, 1.0, 1.0},
{-1.0, 1.0, 1.0},
{-1.0, -1.0, 1.0}, /* left */
{-1.0, 1.0, 1.0},
{-1.0, 1.0, -1.0},
{-1.0, -1.0, -1.0},
{-1.0, -1.0, -1.0}, /* back */
{-1.0, 1.0, -1.0},
{ 1.0, 1.0, -1.0},
{ 1.0, -1.0, -1.0},
324
007-1680-100
pfGeoArray (Geometry Array)
{
{
{
{
1.0, -1.0, 1.0}, /* right */
1.0, -1.0, -1.0},
1.0, 1.0, -1.0},
1.0, 1.0, 1.0},
{-1.0,
{ 1.0,
{ 1.0,
{-1.0,
{-1.0,
{-1.0,
{ 1.0,
{ 1.0,
1.0, 1.0}, /* top */
1.0, 1.0},
1.0, -1.0},
1.0, -1.0},
-1.0, 1.0}, /* bottom */
-1.0, -1.0},
-1.0, -1.0},
-1.0, 1.0}
};
static pfVec3 norms[] =
{
{ 0.0, 0.0, 1.0},
{ 0.0, 0.0, 1.0},
{ 0.0, 0.0, 1.0},
{ 0.0, 0.0, 1.0},
007-1680-100
{-1.0,
{-1.0,
{-1.0,
{-1.0,
0.0,
0.0,
0.0,
0.0,
0.0},
0.0},
0.0},
0.0},
{
{
{
{
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
-1.0},
-1.0},
-1.0},
-1.0},
{
{
{
{
1.0,
1.0,
1.0,
1.0,
0.0,
0.0,
0.0,
0.0,
0.0},
0.0},
0.0},
0.0},
{
{
{
{
0.0,
0.0,
0.0,
0.0,
1.0,
1.0,
1.0,
1.0,
0.0},
0.0},
0.0},
0.0},
{ 0.0, -1.0,
{ 0.0, -1.0,
0.0},
0.0},
325
8: Geometry
{ 0.0, -1.0,
{ 0.0, -1.0,
0.0},
0.0}
};
/* Convert static data to pfMalloc'ed data */
static void*
memdup(void *mem, size_t bytes, void *arena)
{
void *data = pfMalloc(bytes, arena);
memcpy(data, mem, bytes);
return data;
}
/* Set up a PFGS_QUADS pfGeoArray */
garray = pfNewGArray(NULL);
pfGSetPrimType(garray, PFGS_QUADS);
pfGSetNumPrims(garray, 6);
pfGArraySetAttr(garray, PFGA_COORD_ARRAY, 3, GL_FLOAT, 0,
memdup(coords, sizeof(coords), NULL));
pfGArraySetAttr(garray, PFGA_NORMAL_ARRAY, 3, GL_FLOAT, 0,
memdup(norms, sizeof(norms), NULL));
static pfVec3 coords[] =
{
{-1.0, -1.0, 1.0}, /* front */
{ 1.0, -1.0, 1.0},
{ 1.0, 1.0, 1.0},
{-1.0, 1.0, 1.0},
{-1.0, -1.0, 1.0}, /* left */
{-1.0, 1.0, 1.0},
{-1.0, 1.0, -1.0},
{-1.0, -1.0, -1.0},
{-1.0, -1.0, -1.0}, /* back */
{-1.0, 1.0, -1.0},
{ 1.0, 1.0, -1.0},
{ 1.0, -1.0, -1.0},
{ 1.0, -1.0, 1.0}, /* right */
{ 1.0, -1.0, -1.0},
{ 1.0, 1.0, -1.0},
326
007-1680-100
pfGeoArray (Geometry Array)
{ 1.0,
1.0,
{-1.0,
{ 1.0,
{ 1.0,
{-1.0,
1.0, 1.0}, /* top */
1.0, 1.0},
1.0, -1.0},
1.0, -1.0},
{-1.0,
{-1.0,
{ 1.0,
{ 1.0,
1.0},
-1.0, 1.0}, /* bottom */
-1.0, -1.0},
-1.0, -1.0},
-1.0, 1.0}
};
static pfVec3 norms[] =
{
{ 0.0, 0.0, 1.0},
{ 0.0, 0.0, 1.0},
{ 0.0, 0.0, 1.0},
{ 0.0, 0.0, 1.0},
{-1.0,
{-1.0,
{-1.0,
{-1.0,
0.0,
0.0,
0.0,
0.0,
0.0},
0.0},
0.0},
0.0},
{
{
{
{
{
{
{
{
0.0,
0.0,
0.0,
0.0,
1.0,
1.0,
1.0,
1.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
-1.0},
-1.0},
-1.0},
-1.0},
0.0},
0.0},
0.0},
0.0},
{
{
{
{
0.0,
0.0,
0.0,
0.0,
1.0,
1.0,
1.0,
1.0,
0.0},
0.0},
0.0},
0.0},
{
{
{
{
0.0,
0.0,
0.0,
0.0,
-1.0,
-1.0,
-1.0,
-1.0,
0.0},
0.0},
0.0},
0.0}
};
007-1680-100
327
8: Geometry
// Convert static data to pfMalloc'ed data
static void*
memdup(void *mem, size_t bytes, void *arena)
{
void *data = pfMalloc(bytes, arena);
memcpy(data, mem, bytes);
return data;
}
/* Set up a PFGS_QUADS pfGeoArray */
garray = new pfGeoArray;
garray->setPrimType(PFGS_QUADS);
garray->setNumPrims(6);
garray->setAttr(PFGA_COORD_ARRAY, 3, GL_FLOAT, 0,
memdup(coords, sizeof(coords), NULL));
garray->setAttr(PFGA_NORMAL_ARRAY, 3, GL_FLOAT, 0,
memdup(norms, sizeof(norms), NULL));
With pfGeoArrays, unlike with pfGeoSets, you cannot index vertex coordinates and
normals separately. This results in bigger memory requirements. The extra storage is
worth the reduced rendering times, though.
Another example of creating pfGeoArrays can be found in following files:
(IRIX and Linux)
/usr/share/Performer/src/pguide/libpfdu/pfdConvertToGeoArrays.C
/usr/share/Performer/src/pguide/libpr/C++/geoArray.C
(Microsoft Windows)
%PFROOT%\Src\pguide\libpfdu\pfdConvertToGeoArrays.cxx
%PFROOT%\Src\pguide\libpr\C++\geoArray.cxx
Converting pfGeoSets to pfGeoArrays
Since using pfGeoArrays can be much faster on some platforms, such as Onyx4 or Prism
systems, you can convert your geometry from pfGeoSets to pfGeoArrays using the
following two functions:
328
•
pfdConvertGeoSetToGeoArray()
•
pfdConvertNodeGeoSetsToGeoArrays()
007-1680-100
Optimizing Geometry for Rendering
The first function converts an individial pfGeoSet into a pfGeoArray. The second
function traverses a pfNode and replaces all its pfGeoSets with pfGeoArrays. The
parameter flags can be set to 0 or to PFD_CONVERT_TO_INDEXED_GEOARRAYS. In the
second case the loader tries to avoid the use of lengths array and it converts strips to
indexed lines or triangles.
Also, it is possible to use the pseudo loader libpfgeoa to convert the geometry from
pfGeoSets to pfGeoArrays during loading. The pseudo loader is used as follows:
perfly file.ext.geoa
The pseudo loader calls pfdConvertNodeGeoSetsToGeoArrays() with the flag
PFD_CONVERT_TO_INDEXED_GEOARRAYS set. You can overwrite this default by setting
the environment variable PFD_CONVERT_TO_INDEXED_GEOARRAYS to 0.
Optimizing Geometry for Rendering
This section describes how you can use three functions to optimize your pfGeosets or
pfGeoArrays for rendering. The following topics are described:
•
“Function pfdMergeGraph()” on page 329
•
“Function pfdStripGraph()” on page 330
•
“Function pfdSpatializeGraph()” on page 331
•
“The Optimization Pipeline” on page 332
•
“Using the libpfgopt Pseudo Loader” on page 333
Function pfdMergeGraph()
The function pfdMergeGraph() gathers all pfGeoSets referenced in the each static
subgraph rooted at a node. It then creates a new subgraph for each that has a minimum
number of pfGeoSets by grouping pfGeoSets that share state and attribute data. The
function only operates on static subgraphs and, thus, will not destroy pfLOD,
pfSequence, or other dynamic structures in a hierarchy. The new graph, which is not
spatialized, is returned to the caller. All functions described in this section can output
geometry as either pfGeoSets or pfGeoArrays and as either indexed or nonindexed. The
default behavior is to output nonindexed pfGeoArrays if any are present and
007-1680-100
329
8: Geometry
nonindexed pfGeoSets, otherwise. This can be controlled by passing in the following as
the flags parameters:
•
PFD_OUTPUT_GEOSETS
•
PFD_OUTPUT_GEOARRAYS
•
PFD_OUTPUT_INDEXED
These flags force the output format to be of the type and format indicated.
The function pfdMergeGraph() returns the root to a new scene graph and does not
modify the graph that is rooted at node. The calling application must delete the previous
graph if it is no longer needed.
Function pfdStripGraph()
The function pfdStripGraph() collects all pfGeoSets in the graph rooted by the node and
modifies each in several possible ways in order to increase performance. This can include
stripping (converting from separate primitives, such as triangles, to their stripped form,
triangle strips) or unstripping geometry, merging stripped geometry, and reordering
primitives. On certain hardware, such as Onyx4 or Prism systems, reordering primitives
can improve performance by taking advantage of hardware vertex caches to decrease the
number of transformed vertices. If you specify an integer length for cacheLength
OpenGL Performer uses the given value as the length of the cache on the target
hardware. If you do not know the cache length, which varies on some platforms, using a
value of 0 triggers a more generic algorithm to provide better results on machines of
varying cache lengths. Controlling the behavior of the operation can be done by passing
in the OR result of the following tokens as the flags parameter:
PFD_STRIP_UNSTRIPPED_PRIMITIVES
This token indicates that any primitives that are not stripped should be
stripped.
PFD_STRIP_STRIPPED_PRIMITIVES
This token indicates that any primitives that are already stripped should
be restripped.
PFD_UNSTRIP_STRIPPED_PRIMITIVES
This token indicates that any primitives that are already stripped should
be unstripped or converted back to a separated form. If this token and
PFD_STRIP_STRIPPED_PRIMITIVES are both passed, the input
geometry is first unstripped and then restripped.
330
007-1680-100
Optimizing Geometry for Rendering
PFD_MERGE_TRISTRIPS
This token indicates that triangle strips should be joined with
degenerate triangles to construct longer strips. This will reduce the
number of primitives in a pfGeoSet and may improve performance by
reducing the number of draw calls needed to render the geometry.
PFD_REORDER_CACHE_REUSE
This token indicates that the primitives should be reordered to improve
cache reuse. On systems with a hardware vertex cache, this may
improve performance and is typically the most effective reordering
strategy.
PFD_REORDER_REDUCE_DEGENERATES
This token indicates that the reordering of primitives should seek to
reduce the number of mergings of degenerate triangles. This is not
typically as effective as reordering to improve cache reuse.
The function pfdStripGraph() uses the same output flags as pfdMergeGraph().
Note: There is no default behavior for pfdStripGraph(); that is, if flags is set to 0, then
the processing options just described are disabled. In this case, the result will be no
change to the input geometry.
Function pfdSpatializeGraph()
The function pfdSpatializeGraph() operates similarily to pfdSpatialize() and
pfdBreakup(). The function pfdSpatializeGraph() first breaks the geometry of the graph
rooted at a node into smaller chunks to improve spatialization and, in effect, culling. This
breakup turns each geode into a subgraph rooted by a pfGroup node with several
pfGeode children. Next, each static subgraph is spatialized using pfSpatialize(). Similar
to pfdMergeGraph(), this operation does not change the dynamic aspects of a scene
graph. Graphs rooted by nodes such as pfSwitch and pfSequence will not be changed by
the operation. The maxStripLength parameter controls the strip length threshold for the
breakup. If this value is set to 0, then no maximum will be used. The octreeLevels
parameter controls the target number of octree levels for the breakup and spatialization
operations. However, depending upon the number of pfGeoSets in the graph, there may
be slightly more or fewer levels in the actual output graph. Like pfdStripGraph(), the
flags parameter is used to pass in the OR result of the following tokens to control the
behavior of the operation:
007-1680-100
331
8: Geometry
PFD_BREAKUP_GEOSETS
This token indicates that the breakup operation should be
performed.
PFD_SPATIALIZE_GRAPH
This token indicates that the spatialization operation
should be performed.
As with pfdStripGraph(), pfdSpatializeGraph() has no default behavior and will
perform no graph modification if no flags are passed in. Like both pfdStripGraph() and
pfdMergeGraph(), the output can be controlled by the same tokens passed through the
flags parameter, and pfdSpatializeGraph() returns the root to a new scene graph.
The Optimization Pipeline
Collectively, pfdMergeGraph(), pfdStripGraph(), and pfdSpatializeGraph() can be run
as a pipeline on an input scene graph to reformat and optimize the graph for best
performance on a given hardware. Additionally, the following compund tokens are
available for setting the flags which can be passed to all three functions to modify
behavior:
PFD_OPTIMIZE_AGGRESSIVE
This token indicates that the operations should perform their most
aggressive optimizations. This includes restripping all primitives,
reordering to improve cache reuse, merging strips, and performing all
spatialization steps. While this may not always be the best set of
operations, it should provide an excellent starting point, especially for
optimizing on newer hardware, such as Onyx4 or Prism systems.
PFD_OPTIMIZE_CONSERVATIVE
This token is used for performing only the safest optimizations. These
optimizations are unlikely to decrease performance and may be useful
for optimizing across several different platforms and generations of
hardware.
PFD_OPTIMIZE_INDEXED_TRIS
On some hardware, indexed triangles may be the fastest geometry
format. For these systems, the PFD_OPTIMIZE_INDEXED_TRIS token
will use aggressive methods but output to an indexed, unstripped, but
reordered format.
These tokens are combinations of the previous tokens and may be used with any of the
previously mention tokens (most notably, the output tokens). All three compound tokens
include PFD_OUTPUT_INDEXED; therefore, if that mode is not desired, an application
332
007-1680-100
Optimizing Geometry for Rendering
could remove it by performing an AND of token and the negation of
PFD_OUTPUT_INDEXED.
The following code sample illustrates using the optimization functions in a pipeline on
an input graph and saving the output to a file:
root = pfdLoadFile(inputFile);
if (root == NULL)
{
pfNotify(PFNFY_FATAL, PFNFY_USAGE,
"Input graph was not found. Quitting.");
}
// First, call merge graph
root = pfdMergeGraph(root, PFD_OPTIMIZE_AGGRESSIVE);
// Now strip each of the merged geoSets
// length of the cache is 12 vertices
root = pfdStripGraph(root, 12, PFD_OPTIMIZE_AGGRESSIVE);
// Finally, spatialize the graph
// Have 3 octree levels, and no maximum strip length
root = pfdSpatializeGraph(root, 3, 0, PFD_OPTIMIZE_AGGRESSIVE);
// Save new scene graph
pfdStoreFile(root, outputFile);
A sample application is located in the following file:
(IRIX and Linux)
/usr/share/Performer/src/pguide/libpf/C++/optimizeGraph.C
(Microsoft Windows)
%PFROOT%\Src\lib\pguide\libpf\C++\optimizeGraph.cxx
Using the libpfgopt Pseudo Loader
As an alternative to the functions described in this chapter, you can use the libpfgopt
pseudo loader to optimize geometry while loading a database file. You can specify the
filename in one of two formats to call the pseudo loader:
007-1680-100
333
8: Geometry
file.ext.gopt
file.ext.parameters.gopt
If the first format is specified, then the optimizer uses a default operation mode, which
consists of aggressive optimization but with no breakup and with output to indexed
geoarrays. This is equivalent to the following specification in the second format:
file.ext.aggressive,nobreakup,geoarray.gopt
In the second format, parameters is a comma separated list of keywords or
keyword=value pairs. The following are valid keywords:
•
aggressive
•
conservative
•
indexedtris
•
join
•
nojoin
•
cachereuse
•
reducedegens
•
noreorder
•
breakup
•
nobreakup
•
spatialize
•
nospatialize
•
geoset
•
geoarray
•
indexed
•
unindexed
Each of these keywords enables or disables a related operation or mode of the
optimization pipeline, which consists of pfdMergeGraph(), pfdStripGraph(), and
pfdSpatializeGraph().
334
007-1680-100
Rendering 3D Text
The following are valid keyword=value entries for parameters:
•
cachelength=int
•
octree=int
•
striplength=int
They set the related values of the corresponding functions to the specified value. For
more information about all the keywords and their functionality, see the
pfdOptimizeGraph man page.
If at least one parameter is specified, then there is no default operations by the optimizer,
except those explicitly enabled. For example, for the following filename specification, the
optimizer will not join triangle strips, breakup geosets, or spatialize the graph:
foo.pfb.cachereuse.gopt
Additionally, since some keywords have opposing behavior, the order they are specified
matters. For example, if the file is specified in the following manner, then the resulting
geometry will be nonindexed:
foo.pfb.indexed,unindexed.gopt
Rendering 3D Text
In addition to the pfGeoSet and pfGeoArray, libpr offers two other primitives which
together are useful for rendering a specific type of geometry—3D characters. See
Chapter 3, “Nodes and Node Types” and the description for pfText nodes for an example
of how to set up the 3D text within the context of libpf.
pfFont
The basic primitive supporting text rendering is the libpr pfFont primitive. A pfFont is
essentially a collection of pfGeoSets in which each pfGeoSet represents one character of
a particular font. pfFont also contain metric data, such as a per-character spacing, the 3D
escapement offset used to increment a text ‘cursor’ after the character has been drawn.
Thus, pfFont maintains all of the information that is necessary to draw any and all valid
characters of a font. However, note that pfFonts are passive and have little functionality
on their own; for example, you cannot draw a pfFont—it simply provides the character
set for the next higher-level text data object, the pfString.
007-1680-100
335
8: Geometry
Table 8-5 lists some routines that are used with a pfFont.
Table 8-5
pfFont Routines
Routine
Description
pfNewFont()
Create a new pfFont.
pfDelete()
Delete a pfFont.
pfFontCharGSet()
Set the pfGeoSet to be used for a specific character of this pfFont.
pfFontCharSpacing()
Set the 3D spacing to be used to update a text cursor after this character
has been rendered.
pfFontMode()
Specify a particular mode for this pfFont.
Valid modes:
PFFONT_CHAR_SPACING—Specify whether to use fixed or variable
spacings for all characters of a pfFont. Possible values are
PFFONT_CHAR_SPACING_FIXED and
PFFONT_CHAR_SPACING_VARIABLE, the latter being the default.
PFFONT_NUM_CHARS—Specify how many characters are in this font.
PFFONT_RETURN_CHAR—Specify the index of the character that is
considered a ‘return’ character and thus relevant to line justification.
pfFontAttr()
Specify a particular attribute of this pfFont.
Valid attributes:
PFFONT_NAME—Name of this font.
PFFONT_GSTATE—pfGeoState to be used when rendering this font.
PFFONT_BBOX—Bounding box that bounds each individual character.
PFFONT_SPACING—Set the overall character spacing if this is a fixed
width font (also the spacing used if one has not been set for a particular
character).
Example 8-1
Loading Characters into a pfFont
/* Setting up a pfFont */
pfFont *ReadFont(void)
{
pfFont *fnt = pfNewFont(pfGetSharedArena());
for(i=0;i<numCharacters;i++)
{
pfGeoSet* gset = getCharGSet(i);
pfVec3* spacing = getCharSpacing(i);
336
007-1680-100
Rendering 3D Text
pfFontCharGSet(fnt, i, gset);
pfFontCharSpacing(fnt, i, spacing);
}
}
pfString
Simple rendering of 3D text can be done using a pfString. A pfString is an array of font
indices stored as 8-bit bytes, 16-bit shorts, or 32-bit integers. Each element of the array
contains an index to a particular character of a pfFont structure. A pfString can not be
drawn until it has been associated with a pfFont object with a call to pfStringFont(). To
render a pfString once it references a pfFont, call the function pfDrawString().
The pfString class supports the notion of ‘flattening’ to trade off memory for faster
processing time. This causes individual, noninstanced geometry to be used for each
character, eliminating the cost of translating the text cursor between each character when
drawing the pfString.
Example 8-2 illustrates how to set up and draw a pfString.
Example 8-2
Setting Up and Drawing a pfString
/* Create a string a rotate it for 2.5 seconds */
void
LoadAndDrawString(const char *text)
{
pfFont *myfont = ReadMyFont();
pfString *str = pfNewString(NULL);
pfMatrix mat;
float start,t;
/* Use myfont as the 3-d font for this string */
pfStringFont(str, fnt);
/* Center String */
pfStringMode(str, PFSTR_JUSTIFY, PFSTR_MIDDLE);
/* Color String is Red */
pfStringColor(str, 1.0f, 0.0f, 0.0f, 1.0f);
/* Set the text of the string */
pfStringString(str, text);
007-1680-100
337
8: Geometry
/* Obtain a transform matrix to place this string */
GetTheMatrixToPlaceTheString(mat);
pfStringMat(str, &mat);
/* optimize for draw time by flattening the transforms */
pfFlattenString(str);
/* Twirl text for 2.5 seconds */
start = pfGetTime();
do
{
pfVec4 clr;
pfSetVec4(clr, 0.0f, 0.0f, 0.0f, 1.0f);
/* Clear the screen to black */
pfClear(PFCL_COLOR|PFCL_DEPTH, clr);
t = (pfGetTime() - start)/2.5f;
t = PF_MIN2(t, 1.0f);
pfMakeRotMat(mat, t * 315.0f, 1.0f, 0.0f, 0.0f);
pfPostRotMat(mat, mat, t * 720.0f, 0.0f, 1.0f, 0.0f);
t *= t;
pfPostTransMat(mat, mat, 0.0f,
150.0f * t + (1.0f - t) * 800.0f, 0.0f);
pfPushMatrix();
pfMultMatrix(mat);
/* DRAW THE INPUT STRING */
pfDrawString(str);
pfPopMatrix();
pfSwapWinBuffers(pfGetCurWin());
} while(t < 2.5f);
}
338
007-1680-100
Rendering 3D Text
Table 8-6 lists the key routines used to manage pfStrings.
Table 8-6
pfString Routines
Routine
Description
pfNewString()
Create a new pfString.
pfDelete()
Delete a pfString.
pfStringFont()
Set the pfFont to use when drawing this pfString.
pfStringString()
Set the character array that this pfString will represent or render.
pfDrawString()
Draw this pfString.
pfFlattenString()
Flatten all positional translations and the current specification matrix
into individual pfGeoSets so that more memory is used, but no matrix
transforms or translates have to be done between each character of the
pfString.
pfStringColor()
Set the color of the pfString.
pfStringMode()
Specify a particular mode for this pfString.
Valid modes:
PFSTR_JUSTIFY — Sets the line justification and has the following
possible values: PFSTR_FIRST or PFSTR_LEFT, PFSTR_MIDDLE or
PFSTR_CENTER, and PFSTR_LAST or PFSTR_RIGHT.
PFSTR_CHAR_SIZE — Sets the number of bytes per character in the
input string and has the following possible values: PFSTR_CHAR,
PFSTR_SHORT, PFSTR_INT.
007-1680-100
pfStringMat()
Specify a transform matrix that affects the entire character string when
the pfString is drawn.
pfStringSpacingScale()
Specify a scale factor for the escapement translations that happen after
each character is drawn. This routine is useful for changing the spacing
between characters and even between lines.
339
Chapter 9
9. Higher-Order Geometric Primitives
OpenGL Performer also supports a large set of higher-order primitives that you can
include in a scene graph. These higher-order primitives extend the simpler geometric
primitives that can be specified using pfGeodes. “Higher-order” means objects other
than sets of triangles, and typically implies an object that is defined mathematically.
Designs produced by CAD systems are defined by mathematically defined surface
representations. By providing direct support for them, OpenGL Performer expands
possible applications from simple walkthrough ability to direct interaction with the
design data base.
OpenGL Performer also provides classes to define discrete curves and discrete surfaces.
The objects are discussed in the following sections:
007-1680-100
•
“Features and Uses of Higher-Order Geometric Primitives” on page 342
•
“Objects Required by Reps” on page 342
•
“Geometric Primitives: The Base Class pfRep and the Application repTest” on
page 346
•
“Planar Curves” on page 348
•
“Spatial Curves” on page 371
•
“Parametric Surfaces” on page 375
•
“Meshes” on page 413
•
“Subdivision Surfaces” on page 420
341
9: Higher-Order Geometric Primitives
Features and Uses of Higher-Order Geometric Primitives
Higher-order geometric primitives, called representations or simply reps, facilitate the
design process by providing a library of useful curves and surfaces that ease interactive
flexibility, accelerate scene-graph transformations, and reduce the memory footprint of
the scene graph. Reps yield these advantages by using parameters to describe objects.
Instead of a collection of vertices, which must be manipulated independently to change
a surface, reps define surfaces in terms of a relatively small set of control parameters; they
are more like pure mathematical objects.
Reps and the Rendering Process
OpenGL Performer allows you to interact with an abstract object (a representation or rep)
and treat rendering as a separate operation. A simple example of a rep is a sphere,
defined by a radius and a center. After defining a sphere, you can implement how it is
rendered in several ways: by tessellating, by a sphere-specific draw routine, or
conceivably by hardware. Member functions of geometric-primitive classes allow you to
implement the most appropriate way of rendering. The fundamental rendering step of
tessellating a representation is discussed in Chapter 11, “Rendering Higher-Order
Primitives: Tessellators”.
Trimmed NURBS
NURBS curves and surfaces are included in the set of OpenGL Performer reps. OpenGL
also has these, but OpenGL Performer NURBS have two advantages:
•
You can maintain topology, so cracks do not appear at the boundaries of adjacent
tessellations when they are drawn.
•
You have better control over tessellation.
See Chapter 10, “Creating and Maintaining Surface Topology”.
Objects Required by Reps
To use reps effectively, you have to understand the OpenGL Performer representations
of geometric points and the transformation matrices that are used by the methods of the
rep classes.
342
007-1680-100
Objects Required by Reps
New Types Required for Reps
The classes pfRVec2, pfRVec3, and pfRVec4 define two-, three-, and four-dimensional
vectors, and include common operations of vector algebra such as addition, scalar
multiplication, cross products, and so on. See the header file
Performer/pr/pfLinMath.h for a list of operations defined for each vector.
The classes pfRVec2, pfRVec3, pfRVec4 and pfRMatrix are composed of either single- or
double-precision values based on the PF_REAL_IS_DOUBLE #define. Currently,
PF_REAL_IS_DOUBLE is set to 0. Hence, all pfRVecs as well as pfRMatrix and the pfReal
type are defined as single-precision elements. It is important to use these new types
(which are essentially #define statemtents) so that you may change to double- or
arbitrary-precision versions of these elements when this functionality is enabled in
OpenGL Performer.
One more type, pfBool, has also been added to the repertoire of types for
OpenGL Performer and it always maps to the value of a 32-bit integer, regardless of the
value of the PF_REAL_IS_DOUBLE #define, which can be found in
Performer/pf.h.
In addition, pfLoop has been defined as a 32-bit integer and can take on one of the
following values:
•
PFLOOP_OPEN
•
PFLOOP_CLOSED
•
PFLOOP_PERIODIC
•
PFLOOP_UNRESOLVED
Classes for Scalar Functions
The pfScalar class is the base class for defining scalar functions; it allows you to
conveniently read and write functions. The class provides a virtual evaluation method.
007-1680-100
343
9: Higher-Order Geometric Primitives
Class Declaration for pfScalar
The class has the following main methods:
class pfScalar : public pfObject
{ public:
// Creating and destroying
pfScalar();
virtual ~pfScalar();
virtual pfReal eval(pfReal u) = 0;
};
The class pfCompositeScalar allows you to define the functional composition of two
pfScalars.
Class Declaration for pfCompositeScalar
The class has the following main methods:
class pfCompositeScalar : public pfScalar
{ public:
// Creating and destroying
pfCompositeScalar( );
pfCompositeScalar(pfScalar *outFun, pfScalar *inFun);
virtual ~pfCompositeScalar();
// Accessor functions
pfScalar *getOutF()
pfScalar *getInF()
void
setOutF(pfScalar *outF);
void
setInF (pfScalar *inF);
pfReal eval(pfReal t);
Main Features of the Methods in pfCompositeScalar
eval()
Returns the value of outF(inF(t)).
Trigonometric Functions
OpenGL Performer provides classes for two trigonometric functions, pfCosScalar and
pfSinScalar. The class declarations are similar to that of pfScalar.
344
007-1680-100
Objects Required by Reps
Polynomials
Polynomials of arbitrary degree are defined by the class pfPolyScalar.
Class Declaration for pfPolyScalar
The class has the following main methods:
class pfPolyScalar : public pfScalar
{
public:
// Creating and destroying
pfPolyScalar( void );
pfPolyScalar( int degree, pfReal* coef);
virtual ~pfPolyScalar();
// Accessor functions
void set( int degree, pfReal* coef);
int getDegree()
pfReal getCoef( int i)
// Evaluators
pfReal eval(pfReal u);
};
Matrix Class: pfRMatrix
Each geometric primitive is defined with respect to its own coordinate system. The
elementary definition of an object gives a particular orientation and location with respect
to the origin. This reference frame can, in turn, be manipulated by a pfDCS to place it in
a scene or manipulate it.
The base class for higher-order primitives has methods that allow you to locate and
orient a primitive with respect to its own reference frame. These methods make insertion
of pfDCS nodes whenever you want to define the location or orientation of an object or
to change the shape of an object unnecessary.
The location is defined by an pfRVec2 or pfRVec3, and the orientation is controlled by a
3 x 3 matrix, held in the class pfRMatrix. If the matrix is not a rotation matrix, you can
change the shape of an object, for example, you can distort a sphere into an ellipsoid.
007-1680-100
345
9: Higher-Order Geometric Primitives
Geometric Primitives: The Base Class pfRep and the Application repTest
pfRep is the base class for higher-order geometric primitives that are stored in an
OpenGL Performer scene graph. A pfRep is derived from a pfGeode and is therefore
always a leaf node. Figure 9-1 shows the hierarchy of classes derived from pfRep.
The following sections discuss the subclasses of pfRep:
•
“Planar Curves” on page 348
•
“Spatial Curves” on page 371
•
“Parametric Surfaces” on page 375
•
“Meshes” on page 413
•
“Subdivision Surfaces” on page 420
To experiment with pfReps, you can use and modify the application repTest in
/usr/share/Performer/src/pguide/libpf/C++/repTest.C on IRIX and
Linux and in %PFROOT%\Src\pguide\libpf\C++\repTest.cxx on Microsoft
Windows. This code provides sample instances of several geometric representations, as
well as the tessellation and OpenGL Performer calls that render the objects. Sample code
from repTest is included with discussions of several of the classes derived from pfRep.
pfRep has methods to orient the object in space, so you do not have to place a pfDCS
node above each pfRep to move it from its default position.
346
007-1680-100
Geometric Primitives: The Base Class pfRep and the Application repTest
pfPieceWisePolyCurve2d
pfLine2d
pfCircle2d
pfCurve2d
pfSuperQuadCurve2d
pfHsplineCurve2d
pfDisCurve2d
pfNurbCurve2d
pfPieceWisePolyCurve3d
pfLine3d
pfOrientedLine3d
pfCurve3d
pfCircle3d
pfSuperQuadCurve3d
pfHsplineCurve3d
pfGeode
pfNurbCurve3d
pfRep
pfDisCurve3d
pfCompositeCurve3d
pfCuboid
pfPieceWisePolySurface
pfPlane
pfSphere
pfCylinder
pfTorus
pfCone
pfParaSurface
pfSweptSurface
pfRuled
pfCoons
pfNurbSurface
pfHsplineSurface
Figure 9-1
007-1680-100
Class Hierarchy for Higher-Order Primitives
347
9: Higher-Order Geometric Primitives
Class Declaration for pfRep
The class has the following main methods:
class pfRep : public pfGeode
{
public:
pfRep( );
virtual ~pfRep( );
// Accessor functions
void setOrigin( const pfRVec3& org );
void setOrient( const pfRMatrix& mat );
void getOrigin(pfRMatrix& mat );
void getOrient(pfRVec3& org );
};
Main Features of the Methods in pfRep
setOrient()
Sets the orientation of the representation with respect to the origin using
matrix multiplication.
setOrigin()
Sets the location of the representation with respect to the origin. For
example, supplying the vector (1,0,0) shifts the location of the object 1
unit in the direction of the positive X axis.
pfRep’s subclasses typically include evaluator methods to determine coordinates of
points, tangents, and normals. If you do not want the values corresponding to the default
position, do not call these methods before you use setOrient() and setOrigin() to locate
an pfRep. Thus, for example, when defining points on a circle, first set the center and the
radius, then call setOrient() to set the orientation, and then evaluate points.
Planar Curves
A parametric curve in the plane can be thought of as the result of taking a piece of the
real number line, twisting it, stretching it, maybe gluing the ends together, and laying it
down on the plane. The base class for parametric curves that lie in the XY plane is the
class pfCurve2d.
348
007-1680-100
Planar Curves
An important use of pfCurve2d is to specify trim curves, which define boundaries for
surfaces. Surfaces are parameterized by part of a plane, which in OpenGL Performer is
referred to as the uv plane. When an pfCurve2d is used to define a trim curve, it is treated
as a curve in the UV plane. This topic is discussed further in the section “Parametric
Surfaces” on page 375.
Another important use of pfCurve2d is for specifying cross sections for swept surfaces.
See “Swept Surfaces” on page 395.
OpenGL Performer also provides a class to create discrete curves, pfDisCurve2d.
The following sections discuss planar curve classes, most of which are derived from
pfCurve2d:
•
“Mathematical Description of a Planar Curve” on page 349
•
“Lines in the Plane” on page 353
•
“Circles in the Plane” on page 354
•
“Superquadric Curves: pfSuperQuadCurve2d” on page 356
•
“Hermite-Spline Curves in the Plane” on page 359
•
“NURBS Overview” on page 360
•
“NURBS Curves in the Plane” on page 365
•
“Piecewise Polynomial Curves: pfPieceWisePolyCurve2d” on page 367
•
“Discrete Curves in the Plane” on page 368
Mathematical Description of a Planar Curve
Planar curves consist of sets of points, described by two-dimensional vectors, pfRVec2s.
They are parameterized by the pfReal variable t; as t varies, a point “moves” along the
curve. t can be thought of as the amount of time that has passed as a point moves along
the curve. Or, t can measure the distance traveled.
More precisely, each component of a point on the curve is a function of t, which lies in the
parameter interval (t0, t1) on the real line. Points on the curve are described by a pair of
functions of t: (x(t), y(t)).
007-1680-100
349
9: Higher-Order Geometric Primitives
y
t=1.0
t=0.0
x
Object space
0.0
1.0
t
Parameter space
Figure 9-2
Parametric Curve: Parameter Interval (0,1).
Classes derived from pfCurve2d inherit a set of evaluator functions which, for a given
value of t, evaluate a point on the curve, the tangent and normal vectors at the point, and
the curvature. Naturally, the base-class evaluator that locates points on the curve is a
pure virtual function.
To evaluate tangent and normal vectors at a point, pfCurve2d provides virtual functions
that, by default, use finite-central-difference calculations. To compute the tangent to the
curve at p[t], a point on the curve, the tangent evaluator function takes the vector
connecting two “nearby” points on the curve, p[t+∆t] − p[t−∆t] where ∆t is “small,” and
divides by 2∆t. Similarly, a finite-central-difference calculation of the normal vector uses
the difference between two nearby tangent vectors: n[t] = (t[t+∆t] −t[t−∆t])/2∆t.
350
007-1680-100
Planar Curves
Class Declaration for pfCurve2d
The class has the following main methods:
class pfCurve2d : public pfRep
{
public:
// Creating and destroying
pfCurve2d( );
pfCurve2d( pfReal beginT, pfReal endT
);
virtual ~pfCurve2d();
// Accessor functions
void setBeginT( const pfReal beginT );
void setEndT( const pfReal endT );
pfReal getBeginT();
pfReal getEndT();
pfRVec2 getBeginPt();
pfRVec2 getEndPt();
pfRVec2 getBeginTan();
pfRVec2 getEndTan();
void setClosed( const pfLoop loopVal );
pfLoop getClosed();
void setClosedTol( const pfReal tol );
pfReal getClosedTol() const;
// Evaluators
virtual void evalPt(
virtual void evalTan(
virtual void evalNorm(
virtual void evalCurv(
virtual void eval(
pfReal
pfReal
pfReal
pfReal
pfReal
t,
t,
t,
t,
t,
pfRVec2 &pnt ) = 0;
pfRVec2 &tan );
pfRVec2 &norm );
pfReal *curv );
pfRVec2 &pnt,
pfRVec2 &tan,
pfReal *curv,
pfRVec2 &norm );
};
007-1680-100
351
9: Higher-Order Geometric Primitives
Main Features of the Methods in pfCurve2d
pfCurve2d(beginT, endT)
Creates an instance of pfCurve2d(). If you do not specify any arguments,
then the parametric range of the curve is [0.0,1.0].
eval()
For a given t, returns the position, tangent, curvature, and normal
vectors.
evalPt()
Is a pure virtual function to evaluate position on the curve.
evalTan(), evalCurv(), and evalNorm()
Evaluate the curve’s tangent, curvature, and normal vectors,
respectively. The default methods approximate the computation using
central differences taken about a small ∆t, given by (endT - beginT) *
functionTol. functionTol is a static data element specified in the file
pfRep.h.
setBeginT() and setEndT(), getBeginPt() and getEndPt()
Set and get the parameter range for the curve. Whenever you set one of
these values, the endpoint of the curve changes. Therefore, each of these
methods also recomputes the endpoint, which is cached because it is
frequently used. Also, the methods recompute the ∆t used to
approximate derivatives.
Note that all planar curve classes derived from pfCurve2d reuse
setBeginT() and setEndT() to define the extents of their curves.
setClosed() and getClosed()
Set and get whether a curve is closed.
A closed curve is one for which the endpoints match. OpenGL
Performer determines automatically whether curves are closed, but you
can override this with setClosed().
setClosedTol() and getClosedTol()
Set and get the mismatch between endpoints that is allowed when
calculating whether curves are closed.
To specify the origin used to locate an pfCurve2d, use the first two components set by the
inherited method pfRep::setOrigin().
352
007-1680-100
Planar Curves
Lines in the Plane
Parametric lines in the plane are defined by beginning and ending points. The
parameterization is such that as t varies from t1 to t2, a point on the line “moves,” at a
uniform rate, from the beginning to the ending point.
y
(x2,y2)
t=t2
(x1,y1)
t=t1
x
Figure 9-3
Line in the Plane Parameterization
Class Declaration for pfLine2d
The class has the following main methods:
class pfLine2d : public pfCurve2d
{
public:
// Creating and destroying
pfLine2d();
pfLine2d( pfReal x1, pfReal y1, pfReal t1,
pfReal x2, pfReal y2, pfReal t2 );
virtual ~pfLine2d();
// Accessor functions
void setPoint1( pfReal x1, pfReal y1, pfReal t1 ) ;
void setPoint2( pfReal x2, pfReal y2, pfReal t2 ) ;
void getPoint1( pfReal *x1, pfReal *y1, pfReal *t1 ) const;
void getPoint2( pfReal *x2, pfReal *y2, pfReal *t2 ) const;
// Evaluators
void evalPt(
pfReal t, pfRVec2 &pnt );
};
007-1680-100
353
9: Higher-Order Geometric Primitives
Main Features of Methods in pfLine2d
Creates a parametric line with end points (0,0) and (1,0), and parameter
interval (0,1).
pfLine2d()
pfLine2d(x1, y1, t1, x2, y2, t2)
Creates a parametric line starting at the point (x1, y1) and ending at
(x2,y2). The line is parameterized so that t = t1 corresponds to (x1, y1)
and t = t2 corresponds to (x2,y2).
Is the only evaluator function defined for this object. The tangent vector
is (x2-x1, y2-y1) and the curvature is zero.
evalPt()
setPoint*() and getPoint*()
Set and get the end points of the line.
Circles in the Plane
Use the class pfCircle2d to define a parametric circle in the plane. The parameterization
is such that t is the angular displacement, in radians, in a counterclockwise direction
from the X axis. Figure 9-4 illustrates the parameterization of the circle.
y
s
diu
Ra
t
x
Origin
Figure 9-4
354
Circle in the Plane Parameterization
007-1680-100
Planar Curves
Class Declaration for pfCircle2d
The class has the following main methods:
class pfCircle2d : public pfCurve2d
{
public:
// Creating amd destroying
pfCircle2d();
pfCircle2d( pfReal rad, pfRVec2 *org );
virtual ~pfCircle2d();
// Accessor functions
void
setRadius( pfReal rad ) ;
pfReal getRadius() const;
// Evaluator
void evalPt(
pfReal t, pfRVec2 &pnt );
void evalTan( pfReal t, pfRVec2 &tan );
void evalCurv( pfReal t, pfReal *curv );
void evalNorm( pfReal t, pfRVec2 &norm );
void eval(
pfReal t,
pfRVec2 &pnt,
pfRVec2 &tan,
pfReal *curv,
pfRVec2& norm );
};
Main Features of the Methods in pfCircle2d
pfCircle2d inherits methods to set the range of parameter values from pfCurve2d.
pfCircle2d(rad, org)
Creates an instance of a two-dimensional circle with radius rad centered
at org. The default circle has unit radius and origin (0,0). To change the
default position, use the methods setOrigin() and setOrient() inherited
from pfRep.
setRadius() and getRadius()
Set and get the circle’s radius.
pfCircle2d provides exact calculations for the evaluator functions inherited from
pfCurve2d.
007-1680-100
355
9: Higher-Order Geometric Primitives
Superquadric Curves: pfSuperQuadCurve2d
The class pfSuperQuadCurve2d provides methods to define a generalization of a circle
that, when used for constructing a swept surface, is convenient for generating rounded,
nearly square surfaces, or surfaces with sharp cusps (see “Swept Surfaces” on page 395).
Two examples of superquadrics appear in repTest.
The position along the curve is specified by an angle from the x axis, in the same as for
an pfCircle2d. The shape of the curve is controlled by a second parameter.
A superquadric is the set of points (x,y) given by the following equation that clearly
expresses the relationship to the equation of a circle:
( x2 )1 / α + ( y2 )1 / α = ( r2 )1 / α
The above equation can be written in a parametric form:
α
x ( t ) = r cos [ t ] sign [ cos [ t ] ]
α
y ( t ) = r sin [ t ] sign [ sin [ t ] ]
The family of curves generated by these equations as the quantity α varies can be
described as follows (see Figure 9-5).
Four points are always on the curve for any value of α: (±r, 0) and (0, ±r).
356
•
If α is 1, the curve is a circle of radius r.
•
As α approaches zero, the circle expands to fill a square of side 2r as if you were
inflating a balloon in a box.
•
As α approaches infinity, the circle contracts towards the two diameters along the x
and y axes, approaching two orthogonal lines as if you deflated a balloon with two
rigid orthogonal sticks inside it.
007-1680-100
Planar Curves
y
x
Figure 9-5
Superquadric Curve’s Dependence on the Parameter α.
Class Declaration for pfSuperQuadCurve2d
The class has the following main methods:
class pfSuperQuadCurve2d : public pfCurve2d
{
public:
// Creating and destroying
pfSuperQuadCurve2d();
pfSuperQuadCurve2d( pfReal radius,
pfRVec2 *origin,
pfReal exponent );
virtual ~pfSuperQuadCurve2d();
// Accessor functions
void
setRadius( pfReal _radius );
007-1680-100
357
9: Higher-Order Geometric Primitives
pfReal getRadius() const;
void
setExponent( pfReal _exponent );
pfReal getExponent() const;
// Evaluator
void evalPt(
};
pfReal t, pfRVec2 &pnt );
Main Features of the Methods in pfSuperQuadCurve2d
The accessor functions allow you to control the radius r and exponent α of the curve. To
change the default position, use the methods setOrigin() and setOrient() inherited from
pfRep.
358
007-1680-100
Planar Curves
Hermite-Spline Curves in the Plane
A spline is a mathematical technique for generating a single geometric object from pieces.
An advantage of breaking a curve into pieces is greater flexibility when you have many
points controlling the shape: changes to one piece of the curve do not have significant
effects on remote pieces. To define a spline curve for a range of values for the parameter
t, say from 0 to 3, you “tie” together pieces of curves defined over intervals of values for
t. For example, you might assign curve pieces to the three intervals 0 to 1, 1 to 2, and 2 to
3. The four points in the set of parameters, 0, 1, 2, and 3, define the endpoints of the
intervals and are called knots.
A Hermite-spline curve is a curve whose segments are cubic polynomials of the parameter
t, where the coefficients of the polynomials are determined by the position and tangent
to the curve at each knot point. Thus the curve passes through each of a set of specified
points with a specified tangent vector. The set of knot points must be increasing values
of the parameter t.
y
p3
tng0
p1
t=t1
p0
t=t0
tng1
p2
t=t2
t=t3
tng3
tng2
x
Figure 9-6
Hermite Spline Curve Parameterization
Class Declaration for pfHsplineCurve2d
The class for creating Hermite spline curves is pfHsplineCurve2d. The class has the
following main methods:
class pfHsplineCurve2d : public pfCurve2d
{
public:
// Creating and destroying
pfHsplineCurve2d();
virtual ~pfHsplineCurve2d();
007-1680-100
359
9: Higher-Order Geometric Primitives
// Accessor functions
void setPoint(
int i, const pfRVec2 &p );
void setTangent( int i, const pfRVec2 &tng );
void setKnot(
int i, pfReal t );
int getKnotCount() const;
pfRVec2* getPoint( int i );
pfRVec2* getTangent( int i );
pfReal getKnot( int i );
// Evaluator
virtual void evalPt( pfReal t, pfRVec2 &pnt );
};
NURBS Overview
The acronym NURBS stands for “nonuniform rational B-splines.” NURBS define a set of
curves and surfaces that generalizes Bezier curves. Both NURBS curves and Bezier
curves are “smooth” curves that are well suited for CAD design work. They are
essentially determined by a set of points that controls the shape of the curves, although
the points do not lie on the curves.
Because NURBS properties are not widely known, a discussion of their features precedes
details of how to create instances of them. The discussion is necessarily brief and is
intended to provide the minimum amount of information needed to start using OpenGL
Performer NURBS classes.
This general discussion of NURBS is presented in the following sections:
360
•
“OpenGL Performer NURBS Classes” on page 361
•
“NURBS Elements That Determine the Control Parameters” on page 361
•
“Knot Points” on page 362
•
“Control Hull” on page 362
•
“Features of NURBS and Bezier Curves” on page 363
•
“Weights for Control Points” on page 362
007-1680-100
Planar Curves
OpenGL Performer NURBS Classes
The OpenGL Performer classes allow you to treat a NURBS object as a black box that
takes a set of control parameters and generates a geometric shape. A NURBS object’s
essential properties are rather straightforward, although the underlying mathematics are
complex. Unlike lines and circles, NURBS can represent a large set of distinct complex
shapes. Because of this flexibility, developing a NURBS object is often best done
interactively. For example, you could allow a user to design a curve using an interface in
which control parameters are changed by clicking and dragging and by using sliders.
There are three classes:
•
The pfNurbCurve2d class generates curves in the plane, the simplest NURBS object
provided by OpenGL Performer.
•
The pfNurbCurve3d class generates NURBS curves in three-dimensional space.
•
The pfNurbSurface class generates NURBS surfaces, which extend the ideas
underlying NURBS curves to two-dimensional objects. The principles for
controlling the shapes of these objects are all essentially the same.
NURBS Elements That Determine the Control Parameters
This section provides some theoretical background information on NURBS elements. If
you already understand NURBS, continue with “NURBS Curves in the Plane” on
page 365)
NURBS are defined by the following elements, discussed in this chapter:
007-1680-100
•
Nonuniform knot points (see “Knot Points” on page 362)
•
A control hull consisting of control points (see “Control Hull” on page 362)
•
Weighting parameters for control points (see “Weights for Control Points” on
page 362)
361
9: Higher-Order Geometric Primitives
Knot Points
The knot points determine how and where the pieces of the NURBS object are joined. The
knots are nondecreasing— but not necessarily uniformly spaced or distinct—values of
the parameter t for the curve. The sequence of knots need not have uniform spacing in
the parameter interval. In fact, the mathematics of NURBS make it possible and, perhaps,
necessary to repeat knot values; that is, knots can appear with a certain multiplicity. The
number of knot points is determined by counting all the knot points, including all
multiplicities.
For example, although the sequence (0,0,0,0,1,1,1,1) has only two distinct knot points, the
number of knot points is eight. (This example it is the set of knot points for a cubic Bezier
curve defined on the interval 0 to 1). How to determine the order of a NURBS curve is
discussed in “Features of NURBS and Bezier Curves” on page 363.
Control Hull
The control hull is the set of all points that determine the basic shape of NURBS object.
The effect of the control hull is determined by a “B-spline.”
A B-spline is a basis spline; a set of special curves associated with a given knot sequence
from which you can generate all other spline curves having the same knot sequence and
control hull. For each interval described by the knot sequence, the corresponding piece
of a B-spline curve is a Bezier curve.
B-spline curves are like Bezier curves in that they are defined by an algorithm that acts
on a sequence of control points, the control hull, which lie in the plane or in
three-dimensional space.
Weights for Control Points
The third set of control parameters for a NURBS curve is the set of weights associated
with the control points.
A rational B-spline consists of curves that have a weight associated with each control
point. The individual pieces of a NURBS curve usually are not Bezier curves but rational
Bezier curves. The values of the weights have no absolute meaning; they control how
“hard” an individual control point pulls on the curve relative to other control points. If
the weights for all the control points of a rational Bezier curve are equal, then the curve
362
007-1680-100
Planar Curves
becomes a simple Bezier curve. Weights allow construction of exact conic sections, which
cannot be made with simple Bezier curves.
Features of NURBS and Bezier Curves
Bezier curves have the following properties:
•
They are “nice” polynomial curves whose degree is one less than the number of
control points.
For a polynomial curve, each of the components is a polynomial function of the
parameter t. The number of coefficients in the polynomial, the order of the
polynomial, is equal to the number of control points.
•
The control points determine the shape of the Bezier curve, but they do not lie on
the curve, except the first and last control points.
NURBS curves differ in the following ways:
•
The order of the polynomial pieces that make up the NURBS curve depends on the
number of control points and the number of knot points. The order of a NURBS
curve is the difference between the number of knots, accounting for multiplicity,
and the number of control points. That is,
order = number of knot points - number of control points
•
007-1680-100
The relationship between the curves and the control points is looser than for a
Bezier curve. It also depends on the knot sequence and the sequence of weights.
363
9: Higher-Order Geometric Primitives
Equation Used to Calculate a NURBS Curve
The equation that defines the NURBS curve is
∑ B (t)C
p(t) = ---------------------------∑ B (t)W
n
i
i
n
i
i
i
i
•
p(t) is a point on the surface p(t)
•
B in(t) is the ith B-spline basis function of degree n
•
C i is a control point
•
W i is the weight for the control point
Alternative Equation for a NURBS Curve
If you have a surface developed from the alternative expression for a NURBS surface:
∑ B (u)W C
p(u, v) = ----------------------------------∑ B (u)W
n
i
i
i
i
n
i
i
i
you must change the coordinates of the control points to get the same surface from
OpenGL Performer; you convert the coordinates of the control points from (x,y,w) to
(wx,wy,w).
364
007-1680-100
Planar Curves
NURBS Curves in the Plane
The class pfNurbCurve2d defines a nonuniform rational B-spline curve in the plane, the
simplest NURBS object provided by OpenGL Performer.
Class Declaration for pfNurbCurve2d
The class has the following main methods:
class pfNurbCurve2d : public pfCurve2d
{
public:
// Creating and destroying
pfNurbCurve2d( );
pfNurbCurve2d( pfReal tBegin, pfReal tEnd );
virtual ~pfNurbCurve2d( );
// Accessor functions
void setControlHull(
void setControlHull(
void setWeight(
void setKnot(
void setControlHullSize(
int
int
int
int
int
i,
i,
i,
i,
s
const pfRVec2 &p );
const pfRVec3 &p );
pfReal w );
pfReal t );
);
pfRVec2* getControlHull( int i );
pfReal getWeight( int i );
int
getControlHullSize( );
int
getKnotCount( );
pfReal getKnot( int i );
int
getOrder( );
void removeControlHullPnt(int i);
void removeKnot(int i);
// Evaluator
virtual void evalPt( pfReal t, pfRVec2 &pnt );
};
Main Features of the Methods in pfNurbCurve2d
pfNurbCurve2d(tBegin, tEnd)
Creates a NURBS curve in the plane with the specified parameter
domain. The default parameter domain is 0.0 to 1.0.
007-1680-100
365
9: Higher-Order Geometric Primitives
evalPt()
Is a pure virtual function inherited from pfCurve2d, and produces
unpredictable results until you set the control parameters.
setControlHull(i, p) and getControlHull(i)
Set and get the two-dimensional control point with index i to the value
p. If you supply pfRVec3 arguments, the location of the control points is
set by the first two components; the last component is their weight.
setControlHullSize()
Gives a hint about how big the control hull array is. This is not
mandatory but uses time and space most efficiently.
setKnot(i, t) and getKnot(i)
Set and get the knot point with index i and the value t.
setWeight(i, w) and getWeight(i)
Set and get the weight of the control point with index i and weight w.
366
007-1680-100
Planar Curves
Piecewise Polynomial Curves: pfPieceWisePolyCurve2d
A piecewise polynomial curve consists of an array of polynomial curves. Each
polynomial curve is a polynomial mapping from t to UV plane, where the domain is a
subinterval of [0,1]. The polynomial coefficients are set by setControlHull().
Notice that an pfPieceWisePolyCurve2d is a subclass of pfCurve2d. The domain of a
pfPieceWisePolyCurve2d is defined to be [0, n] where n is the number of pieces.
If reverse is 0, then for any given t in [0, n], its corresponding uv is evaluated in the
following way: The index of the piece that corresponds to t is floor(t), and the polynomial
of that piece is evaluated at w1 + (t-floor(t)) * (w2-w1) to get the (u,v), where [w1, w2] is the
domain interval (set by setLimitParas()) of this piece.
If reverse is 1, then for any given t in [0,n], we first transform t into n-t, then perform
the normal evaluation (at n-t) as described in the preceding paragraph.
Class Declaration for pfPieceWisePolyCurve
The class has the following main methods:
class pfPieceWisePolyCurve2d : public pfCurve3d
{
public:
// Creating and destroying
pfPieceWisePolyCurve2d ( );
virtual ~pfPieceWisePolyCurve2d ( );
//Accessor functions
void setControlHull ( int piece, int i, const pfRVec2& p);
pfRVec2& getControlHull ( int piece, int i);
void setLimitParas ( int piece, pfReal w1, pfReal w2);
void setReverse ( int _reverse);
pfRVec2& getLimitParas ( int piece);
int getReverse ( ) const;
int getPatchCount ( ) const;
int getOrder ( int piece);
virtual void evalPt ( pfReal t, pfRVec2& pnt);
virtual void evalBreakPoints ( pfParaSurface* sur);
};
007-1680-100
367
9: Higher-Order Geometric Primitives
setControlHull(piece, i, p) defines the ith polynomial coefficient of the pieceth polynomial
curve to p. p[0] is for the u coefficient and p[1] is for the v coefficient. setLimitParas() sets
the domain interval.
The class pfPieceWisePolyCurve3d has parallel functionality and declaration.
Discrete Curves in the Plane
The class pfDisCurve2d is the base class for making a discrete curve from line segments
connecting a sequence of points in the plane. Because pfDisCurve2d is not derived from
pfCurve2d, it does not inherit that class’s finite difference functions for calculating
derivatives, therefore, pfDisCurve2d includes member functions that calculate arc
length, tangents, principal normals, and curvatures using finite central differences.
Figure 9-7 illustrates the definition of the curve by a set of points.
y
pi=(points[2i], points[2i+1])
p3
p1
p4
p2
p0
x
Figure 9-7
368
Discrete Curve Definition
007-1680-100
Planar Curves
Class Declaration for pfDisCurve2d
The class has the following main methods:
class pfDisCurve2d : public pfRep
{
public:
// Creating and destroying
pfDisCurve2d( void );
pfDisCurve2d( int nPoints, pfReal *points );
virtual ~pfDisCurve2d( void );
// Accessor functions
void set (int nPoints, pfReal* points);
pfRVec2 getBeginPt() const;
pfRVec2 getEndPt() const;
pfLoop getClosed();
void
setClosed( pfLoop c );
void setPoint( int i, const pfRVec2& pnt );
pfRVec2 getPoint( int i) const;
int getPointCount(); const;
pfRVec2 getTangent(int i) const;
pfRVec2 getNormal(int i) const;
pfReal getCurvature(int i) const;
// Evaluators
void computeTangents( );
void computeNormals( );
void computeCurvatures( );
void computeDerivatives( );
};
007-1680-100
369
9: Higher-Order Geometric Primitives
Main Features of the Methods in pfDisCurve2d
pfDisCurve2d(nPoints, points)
Creates a discrete curve from an array of point coordinates. The
constructor assumes that the coordinates of the points are stored in pairs
sequentially; thus the points array is nPoint*2 in length.
computeCurvatures()
Computes the curvature, which is the magnitude of the normal vector.
computeDerivatives()
Is a convenience function that calls (in order) the tangent, normal, and
curvature functions.
computeNormals()
Computes the principal normal at a point using finite central differences
and stores the result in the class member pfDvector n. For the point p[i],
the normal vector is computed to be the difference vector between the
tangents at the two neighboring points, t[i+1] - t[i-1], divided by the sum
of the distances from p[i] to the two neighboring points.
computeTangents()
Computes the arc lengths of segments and then uses finite central
differences to compute the tangents. For the point p[i], the tangent
vector is computed to be the vector between its two neighboring points,
p[i+1] - p[i-1], divided by the sum of the distances from p[i] to the two
neighboring points. The tangents are stored in the pfDvector t, the arc
lengths in the pfDvector ds, and the total arc length in arcLength.
getCurvature() Returns the value of the curvature at the ith point.
getNormal()
Returns the value of the normal at the ith point.
getPoint()
Returns the value of the ith point.
getPointCount()
Returns the value of the ith point.
getTangent()
370
Returns the value of the tangent at the ith point.
007-1680-100
Spatial Curves
Spatial Curves
The class pfCurve3d is the base for parametric curves that lie in three-dimensional space.
Among other uses, a curve in space could locate a moving viewpoint in a CAD
walk-through.
The nature of these curves is essentially the same as those of pfCurve2d curves, except
pfCurve3d curves are made of points described by pfRVec3s. The components of the
points are assumed to be x, y, and z coordinates. Refer to the section “Planar Curves” on
page 348 for a discussion of the basic features of parametric curves.
This section parallels the discussion in “Planar Curves” on page 348, and emphasizes the
(not very great) differences that distinguish spatial curves:
•
“Lines in Space” on page 371
•
“Circles in Space” on page 372
•
“Superquadrics in Space” on page 372
•
“Hermite Spline Curves in Space” on page 373
•
“NURBS Curves in Space” on page 373
•
“Curves on Surfaces: pfCompositeCurve3d” on page 374
•
“Discrete Curves in Space” on page 375
The class declaration for pfCurve3d is in the file
/usr/include/Performer/pf/pfCurve3d.h on IRIX and Linux and
%PFROOT%\Include\pf\pfCurve3d.h on Microsoft Windows. Its declaration is
essentially identical to the declaration for pfCurve2d. The difference is that all pfRVec2
variables are replaced by pfRVec3 variables.
Lines in Space
The base class for lines in space, pfLine3d, is essentially the same as pfLine2d, discussed
in “Lines in the Plane” on page 353. The main differences are due to the need to manage
three-dimensional vectors. Thus all vector variables are pfRVec3 and the constructor
takes six variables to define the endpoints of the line.
007-1680-100
371
9: Higher-Order Geometric Primitives
The default orientation of the curve is identical to that for the planar curve pfLine2d; you
can translate and rotate the line in three-dimensional space with the methods setOrigin()
and setOrient() inherited from pfRep.
pfOrientedLine3d
The class pfOrientedLine3d is derived from pfLine3d, and adds vectors to define a
moving three-dimensional reference frame for the line. This object is useful if you want
a straight-line path for an pfFrenetSweptSurface (see “Swept Surfaces” on page 395 and,
in particular, “Class Declaration for pfFrenetSweptSurface” on page 399).
The methods of pfOrientedLine3d add to the description of the line an “up” vector,
which you specify. The normal to the line is calculated from the direction of the line and
the up vector.
Circles in Space
The class pfCircle3d defines a parametric circle with an arbitrary location and orientation
in space. The parameterization of the circle, before you change its location or orientation,
is such that t is the angular displacement, in radians, in a counterclockwise direction
from the x axis.
The class declaration for pfCircle3d is identical to that for pfCircle2d, discussed in
“Circles in the Plane” on page 354, except for the changes from pfRVec2 to pfRVec3. The
member functions perform the same operations. For more information, see the
discussion in the section “Circles in the Plane” on page 354.
If the matrix you use to orient an pfCircle3d does not correspond to a rotation about an
axis—that is, the matrix is not orthonormal— you not only change the tilt of the plane in
which the circle lies but also change the radius, and may distort the circle into an ellipse. .
Superquadrics in Space
The class pfSuperQuadCurve3d provides methods to define a superquadric in space (see
“Superquadric Curves: pfSuperQuadCurve2d” on page 356). The class declaration is
identical to that for pfSuperQuad2d except that position on the curve is defined by an
pfRVec3.
372
007-1680-100
Spatial Curves
The default orientation of the curve is identical to that for the planar curve
pfSuperQuad2d; you can translate and rotate the curve in three-dimensional space with
the methods setOrigin() and setOrient() inherited from pfRep.
Hermite Spline Curves in Space
The class pfHsplineCurve3d provides methods to define a Hermite spline curve in space.
The definition of the curve is the same as that for a Hermite spline curve in the plane,
discussed in “Hermite-Spline Curves in the Plane” on page 359. The class declaration is
the same as that for pfHsplineCurve2d, but the position and tangent vectors are
pfRVec3s.
NURBS Curves in Space
The basic properties of NURBS are discussed in the section “NURBS Overview” on
page 360. In an effort to keep things as simple as possible, the discussion in that section
has a bias toward curves in the plane. But the principles and control parameters are, with
one difference, the same for NURBS curves in space.
The difference is that control points for NURBS curves in space can be anywhere in space
instead of being restricted to a plane. The section “Examples of NURBS Curves” in
Chapter 8 of The Inventor Mentor presents illustrations of NURBS curves in space, along
with their control parameters.
The class pfNurbCurve3d is the base class for NURBS curves in space. Its class
declaration is practically identical to that for pfNurbCurve2d but all occurrences of
pfRVec2 are changed to pfRVec3. In addition, the vector argument of setControlHull()
can be an pfRVec3, if you just want to specify control point locations, or an pfRVec4, if
you want to append weighting information as a fourth component. See the discussion in
the section “NURBS Curves in the Plane” on page 365.
007-1680-100
373
9: Higher-Order Geometric Primitives
Curves on Surfaces: pfCompositeCurve3d
A planar curve in the u-v plane describes a curve on the surface, given a parameterized
surface (see the section “Parametric Surfaces” on page 375). Each point on the curve in
the parameter plane is “lifted up” to the surface. Such curves are known as composite
curves because they are described mathematically as the composition of the function
describing the curve and the function describing the surface. The edge of a surface
defined by a trim curve is a composite curve.
pfCompositeCurve3d is the base class for composite curves. This class is useful for
defining trim curves and surface silhouettes in the parametric surface’s coordinate
system.
Class Declaration for pfCompositeCurve3d
The class has the following main methods:
class pfCompositeCurve3d : public pfCurve3d
{
public:
// Creating and destroying
pfCompositeCurve3d( );
pfCompositeCurve3d( pfParaSurface *sur, pfCurve2d *cur );
virtual ~pfCompositeCurve3d( );
// Accessor functions
void set( pfParaSurface *sur, pfCurve2d *cur );
pfParaSurface* getParaSurface() const;
pfCurve2d* getCurve2d() const;
// Evaluator
void evalPt( pfReal u, pfRVec3 &pnt );
};
Main Features of the Methods in pfCompositeCurve3d
The constructor takes two arguments: the first is the surface on which the curve lies, the
second is the curve in the coordinate system of the surface. The returned object is a curve
in space.
374
007-1680-100
Parametric Surfaces
Discrete Curves in Space
The class pfDisCurve3d is the base class for making a discrete curve of line segments
connecting a sequence of points in space. The class declaration for pfDisCurve3d is
identical to that for pfDisCurve2d, discussed in “Discrete Curves in the Plane” on
page 368, but pfRVec2 changes to pfRVec3. The member functions perform the same
operations.
Example of Using pfDisCurve3d and pfHsplineCurve3d
One application of an pfDisCurve3d and pfHsplineCurve3d is to use them to
interactively specify routing for tubing. These are the operations to perform:
1.
Create a pfDisCurve3d from a set of points. See “Discrete Curves in Space” on
page 375.
2. Use the points and tangents to the discrete curve to create a continuous path with an
pfHsplineCurve3d. See “Hermite Spline Curves in Space” on page 373
3. Use the continuous path in an pfFrenetSweptSurface with a circular cross section.
See “pfFrenetSweptSurface” on page 399.
Parametric Surfaces
A parametric surface can be thought of as the result of taking a piece of a plane, twisting
and stretching it, maybe gluing edges of the piece together, and placing it in space.
The introductory discussion of parametric surfaces occurs in the following sections:
007-1680-100
•
“Mathematical Description of a Parametric Surface” on page 376
•
“Defining Edges of a Parametric Surface: Trim Loops and Curves” on page 377
•
“Adjacency Information: pfEdge” on page 379
•
“Base Class for Parametric Surfaces: pfParaSurface” on page 380
375
9: Higher-Order Geometric Primitives
The subclasses of pfParaSurface are discussed in the subsequent sections:
•
“pfPlaneSurface” on page 384
•
“pfSphereSurface” on page 386
•
“pfCylinderSurface” on page 389
•
“pfTorusSurface” on page 391
•
“pfConeSurface” on page 393
•
“Swept Surfaces” on page 395
•
“Ruled Surfaces” on page 400
•
“Coons Patches” on page 402
•
“NURBS Surfaces” on page 404
•
“Hermite-Spline Surfaces” on page 411
Instances of most of the pfParaSurface subclasses are used in the sample application
/usr/share/Performer/src/pguide/libpf/C++/repTest.C on IRIX and
Linux and %PFROOT\Src\pguide\libpf\C++\repTest.cxx on Microsoft
Windows.
Mathematical Description of a Parametric Surface
To locate a point on a parametric surface, you need two parameters, referred to as u and
v in OpenGL Performer. The set of u and v values that describe the surface are known as
the parameter space, or coordinate system, of the surface (see Figure 9-8).
More precisely, the coordinates of the points in space that define a parametric surface are
described by a set of three functions of two parameters: (x(u,v), y(u,v), z(u,v)).
Well-known examples of a parametric surface are a sphere or a globe. On a globe you can
locate points with two parameters: latitude and longitude. The rectangular grid of
latitude and longitudes is the coordinate system that describes points on the globe.
376
007-1680-100
Parametric Surfaces
z
y
x
v
1.0
0.0
Figure 9-8
u
1.0
Parametric Surface: Unit-Square Coordinate System
Defining Edges of a Parametric Surface: Trim Loops and Curves
To define the extent of a parametric curve, pick an interval. For accurate trimming of a
parametric surface, you need more complex tools. You are likely to need:
•
Edges for the surface other than those defined by the limits of the coordinate
system. For example, to define a pipe elbow, you might join two cylinders by a piece
cut from a torus.
•
Holes in a surface, for example, to define a T-joint intersection of pipes.
OpenGL Performer keeps the trim loop side on the left as you look down on the u-v plane
while a point moves along the curve in the direction of increasing t; you can hold on to
007-1680-100
377
9: Higher-Order Geometric Primitives
the surface with your left hand as you go along the trim loop. Thus a clockwise loop
removes a hole; a counterclockwise loop keeps the enclosed region and eliminates
everything outside. Do not create a trim loop that crosses itself like a figure eight.
OpenGL Performer allows you to maintain curves to define the edges of a surface. These
curves are pfCurve2d objects defined in the u-v plane that are “lifted” to the surface by
the parameterization. The main use of these curves is to eliminate a portion of the surface
on one side of the curve. The name of a curve in the coordinate system that is used to
define (possibly a piece of) such a surface edge is a trim curve. One or more joined trim
curves form a sequence called a trim loop. To be of use, trim curves should form a closed
loop or reach the edges of the coordinate system for the surface. Figure 9-9 illustrates
trim loops and their effect on a surface.
z
y
x
Trim1
v
Trim2
Trim3
u
Figure 9-9
378
Trim Loops and Trimmed Surface: Both Trim Loops Made of Four Trim Curves
007-1680-100
Parametric Surfaces
Adjacency Information: pfEdge
An pfEdge defines a trim curve in u, v space. pfEdge holds information about a surface’s
adjacency. Each pfEdge identifies an pfBoundary, which the class pfTopo uses to keep
track of surface connectivity, and continuous and discrete versions of the trim curve
associated with the boundary. The members of an pfEdge are set by the toplogy building
tools; the methods of pfEdge access the members. Topology building and the classes
pfTopo and pfBoundary are discussed further in Chapter 10, “Creating and Maintaining
Surface Topology”.
The information held in pfEdge allows tessellators to determine whether a set of vertices
has already been developed for points shared with other surfaces. You can also find other
surfaces that have the same edge or trim-curve endpoint as that defined by a given trim
curve.
The set*() methods are mainly used when reading surface data from a file and creating
OpenGL Performer data structures.
Class Declaration for pfEdge
The class has the following main methods:
class pfEdge : public pfObject
{
public:
// Creating and destroying
pfEdge();
~pfEdge();
pfCurve2d *getContCurve();
void setContCurve(pfCurve2d *c);
pfDisCurve2d *getDisCurve();
void setDisCurve( pfDisCurve2d *d);
int getBoundary();
void setBoundaryDir( int dir );
int getBoundaryDir();
};
007-1680-100
379
9: Higher-Order Geometric Primitives
Base Class for Parametric Surfaces: pfParaSurface
pfParaSurface is the base class for parametric surfaces in OpenGL Performer. As for the
base classes pfCurve2d and pfCurve3d, pfParaSurface includes a pure virtual function
to evaluate points on the surface and default evaluator functions that calculate
derivatives using finite central differences. The surface normal at a point is the cross
product of the partial derivatives.
For parametric curves whose extent is defined by the interval of values for t, the extent
of an pfParaSurface is, initially, defined by all the points in its parameter space.
Class Declaration for pfParaSurface
The class has the following main methods:
class pfParaSurface : public pfRep
{
public:
// Creating and destroying
pfParaSurface( pfReal _beginU = 0, pfReal _endU = 1,
pfReal _beginV = 0, pfReal _endV = 1,
int
_topoId = 0, int
_solid_id = -1 );
virtual ~pfParaSurface();
// Accessor functions
void setBeginU( pfReal u );
void setEndU(
pfReal u );
void setBeginV( pfReal v );
void setEndV(
pfReal v );
void setSolidId( int solidId);
void SetTopoId( int topoId);
void setSurfaceId (int surfaceId);
pfReal getBeginU() const;
pfReal getEndU() const;
pfReal getBeginV() const;
pfReal getEndV() const;
380
int
pfLoop
int
pfEdge*
getTrimLoopCount();
getTrimLoopClosed( int loopNum );
getTrimCurveCount( int loopNum );
getTrimCurve( int loopNum, int curveNum );
int
getTopoId();
007-1680-100
Parametric Surfaces
int
int
getSolidId();
getSurfaceId();
void setHandednessHint( pfbool _clockWise );
pfbool getHandednessHint() const;
pfGeoState* getGState( ) const;
int setGState( pfGeoState *gState );
void insertTrimCurve( int loopNum, pfCurve2d *c, pfDisCurve2d *d );
// Explicit add a trim curve to a trim loop
void addTrimCurve(int loopNum, pfCurve2d *c, pfDisCurve2d *d );
void setTrimLoopClosed( int loopNum, pfLoop closed );
// Surface evaluators
virtual void evalPt(
virtual void evalDu(
virtual void evalDv(
virtual void evalDuu(
virtual void evalDvv(
virtual void evalDuv(
virtual void evalNorm(
pfReal
pfReal
pfReal
pfReal
pfReal
pfReal
pfReal
u,
u,
u,
u,
u,
u,
u,
pfReal
pfReal
pfReal
pfReal
pfReal
pfReal
pfReal
v,
v,
v,
v,
v,
v,
v,
pfRVec3
pfRVec3
pfRVec3
pfRVec3
pfRVec3
pfRVec3
pfRVec3
&pnt ) = 0;
&Du );
&Dv );
&Duu );
&Dvv );
&Duv );
&norm );
// Directional derivative evaluators
virtual void evalD( pfReal u, pfReal v, pfReal theta, pfRVec3 &D );
virtual void evalDD( pfReal u, pfReal v, pfReal theta, pfRVec3 &DD );
virtual void eval( pfReal u, pfReal v,
pfRVec3 &p,
// The point
pfRVec3 &Du,
// The derivative in the
pfRVec3 &Dv,
// The derivative in the
pfRVec3 &Duu,
// The 2nd derivative in
pfRVec3 &Dvv,
// The 2nd derivative in
pfRVec3 &Duv,
// The cross derivative
pfReal &s,
// Texture coordinates
pfReal &t );
u direction
v direction
the u direction
the v direction
void clearTessallation();
};
007-1680-100
381
9: Higher-Order Geometric Primitives
Main Features of the Methods in pfParaSurface
addTrimCurve(j, curve, discurve)
Is a quick function for building a trim loop that assumes you know the
order of trim curves. It adds curve to the end of the list of continuous trim
curves for the jth trim loop, and adds discurve to the list of discrete trim
curves.
For example, you could build the trim loops in Figure 9-9 by starting
with one segment and successively adding segments. If the beginning
of curve does not match the end of the previously added curve, use
insertTrimCurve(), which finds the right place for the curve by
assuming topological consistency.
eval()
Returns the evaluator functions. The last two arguments of eval() are the
same as the input coordinates u and v.
evalDu(), evalDv(), evalDuu(), evalDvv(), and evalDuv()
Are evaluator functions that use central differences to calculate the first
and second derivatives, identified by the lowercase u and v in the
function names, at a point on the surface.
evalD() and evalDD()
Calculate the first and second directional derivatives in the direction
given by an angle theta from the u axis in the parameter space.
evalNorm()
Calculates the unit normal to the surface.
evalPt()
Is a pure virtual function that you define to specify a surface.
pfParaSurface()
Constructs a parametric surface. You can specify the topology and the
surface to which the parametric surface belongs. See “Summary of Scene
Graph Topology: pfTopo” on page 428.
insertTrimCurve(j, curve, discurve)
Is a slower function than addTrimCurve() for building a trim loop that
attempts to guarantee all curves form a sensible trim loop sequence. It
compares the ends of curve with the ends of the trim curves that are
already in the jth trim loop and inserts curve at the appropriate point in
the list. Similarly, addTrimCurve() inserts the discrete curve discurve. If
insertTrimCurve() cannot find an endpoint match, it adds curve to the
end of the list of trim curves. If you are building a trim loop by inserting
trim curves end to end, then addTrimCurve() gives the same result but
more quickly.
382
007-1680-100
Parametric Surfaces
setBeginU(), setBeginV(), etc.
Set and get the start and end values for the coordinate space of the
surface. The coordinate space is a rectangle in the UV plane. The default
is the unit square; u and v both lie in the interval (0,1).
getTrimLoopCount()
Returns the number of trim loops for the pfParaSurface.
getTrimLoopClosed() and setTrimLoopClosed()
Get and set the flag indicating whether a given trim loop is closed.
OpenGL Performer determines this for you, so use
setTrimLoopClosed() with caution; you could get a meaningless result.
getTrimCurveCount()
Returns the number of trim curves in the specified trim loop.
getTrimCurve(i,j)
Returns the pfEdge for the trim curve with index i in the trim loop with
index j.
clearTessellation()
Removes all data that resulted from previous tessellation. This removal
allows the surface to be retessellated with a different tolerance. For each
trim curve, the disCurve is deleted if the contCurve is not NULL. The
xyzBoundary in its boundary structure is deleted. Also, the tessellated
triangles (csGeometry) are removed.
getGState() and setGState()
Get and set the pfGeoState to be used when tessellating the surface and
setting pfGeoStates on generated geometry (pfGeoSets). setGState()
returns 1 if successful, –1 otherwise.
007-1680-100
383
9: Higher-Order Geometric Primitives
pfPlaneSurface
The simplest parametric surface is a plane. The class pfPlaneSurface defines a plane by
two parameter intervals and three points that define the two coordinate directions.
Figure 9-10 illustrates the parameterization of an pfPlaneSurface.
z
(x3,y3,z3)
u=u1
v=v3
y
(x1,y1,z1)
u=u1
v=v1
(x2,y2,z2)
u=u2
v=v1
x
Figure 9-10
Plane Parameterization
Class Declaration for pfPlaneSurface
The class has the following main methods:
class pfPlaneSurface : public pfParaSurface
{
public:
// Creating and destroying
pfPlaneSurface();
pfPlaneSurface( pfReal x1, pfReal y1, pfReal z1, pfReal u1, pfReal v1,
pfReal x2, pfReal y2, pfReal z2, pfReal u2,
pfReal x3, pfReal y3, pfReal z3, pfReal v3 );
virtual ~pfPlaneSurface();
// Accessor functions
void setPoint1( pfReal x1, pfReal y1, pfReal z1, pfReal u1, pfReal v1);
void setPoint2( pfReal x2, pfReal y2, pfReal z2, pfReal u2 );
void setPoint3( pfReal x3, pfReal y3, pfReal z3, pfReal v3 );
384
007-1680-100
Parametric Surfaces
void getPoint1( pfReal
pfReal
void getPoint2( pfReal
void getPoint3( pfReal
*x1,
*u1,
*x2,
*x3,
pfReal
pfReal
pfReal
pfReal
*y1, pfReal *z1,
*v1 );
*y2, pfReal *z2, pfReal *u2 );
*y3, pfReal *z3, pfReal *v3 );
// Evaluators
void evalPt( pfReal u, pfReal v, pfRVec3 &pnt );
void evalDu( pfReal u, pfReal v, pfRVec3 &Du );
void evalDv( pfReal u, pfReal v, pfRVec3 &Dv );
void evalNorm( pfReal u, pfReal v, pfRVec3 &norm );
}
Main Features of the Methods in pfPlaneSurface
pfPlaneSurface()
When you construct the class, you can specify the plane with three
points and two parameter intervals or you can use the setPoint*()
methods. Those parameters have the following meanings:
•
the point (x1,y1,z1) and its parameter values, (u1,v1)
•
the point (x2,y2,z2), which defines the u direction,
(x2-x1,y2-y1,z2-z1), and its parameter values (u2,v1)
•
the point (x3,y3,z3), which defines the v direction,
(x3-x1,y3-y1,z3-z1) and its parameter values (u1,v3).
setPoint*() and getPoint*()
Set and get each of the points that define the plane and their
corresponding parameter values (see pfPlaneSurface()).
007-1680-100
385
9: Higher-Order Geometric Primitives
pfSphereSurface
The surface of the sphere is parameterized by angles, in radians, for latitude and
longitude; v corresponds to longitude, u to latitude. Figure 9-11 illustrates the
parameterization of an pfSphereSurface.
z
pnt
v
y
Radius
Origin
u
x
Figure 9-11
386
Sphere Parameterization
007-1680-100
Parametric Surfaces
Class Declaration for pfSphereSurface
The class has the following main methods:
class pfSphereSurface : public pfParaSurface
{
public:
// Creating and destroying
pfSphereSurface( );
pfSphereSurface( pfReal radius );
virtual ~pfSphereSurface( );
// Accessor functions
void setRadius( pfReal radiusVal );
pfReal getRadius( ) const;
// Evaluators
void evalPt(
pfReal
void evalNorm( pfReal
}
u, pfReal v, pfRVec3 &pnt );
u, pfReal v, pfRVec3 &norm );
Main Features of the Methods in pfSphereSurface
The constructor defines a sphere centered on the origin with the specified radius. The
default radius is 1. The evaluator functions do not use finite-difference calculations for
derivatives.
Any point on the sphere is represented as:
x = radius * cos(u) * sin(v)
y = radius * sin(u) * sin(v)
z = radius * cos(v)
007-1680-100
387
9: Higher-Order Geometric Primitives
pfSphereSurface Example
The following code from the sample application repTest illustrates how an instance of an
pfSphereSurface of radius three would be created:
pfSphereSurface *sphere = new pfSphereSurface( 3 );
// under certain conditions, a trim curve is added that keeps only the
// portion of the surface above a circle
if ( nVersions <= 0 )
{
pfCircle2d *trimCircle2d =
new pfCircle2d( 1.0, new pfRVec2(M_PI/2.0,M_PI) );
sphere->addTrimCurve( 0, trimCircle2d );
}
setUpShape( sphere, PF_XDIST*numObject++, Y, PF_VIEWDIST );
setUpShape() locates the sphere in the scene, tessellates it, and places it in the scene
graph (see /usr/share/Performer/src/pguide/libpf/C++/repTest.C on
IRIX and Linux and %PFROOT%\Src\pguide\libpf\C++\repTest.cxx on
Microsoft Windows). Creating an instance of any pfRep is basically the same, as
subsequent examples in the discussions of other pfReps will show.
388
007-1680-100
Parametric Surfaces
pfCylinderSurface
The pfCylinderSurface class provides methods for describing a cylinder.
A cylinder can be defined geometrically as the surface in space that is swept by moving
a circle along an axis that is perpendicular to the plane of the circle and passes through
the center of the circle.
The parameterization of an pfCylinderSurface is as follows: u represents the position on
the circle and that v represents the position along the axis.
z
v
pnt
y
Origin
Height
Radiu
s
u
x
Figure 9-12
007-1680-100
Cylinder Parameterization
389
9: Higher-Order Geometric Primitives
Class Declaration for pfCylinderSurface
The class has the following main methods:
class pfCylinderSurface : public pfParaSurface
{
public:
// Creating and destroying
pfCylinderSurface( void );
pfCylinderSurface( pfReal radius, pfReal height );
virtual ~pfCylinderSurface();
// Accessor functions
void setRadius( pfReal radiusVal ) ;
void setHeight( pfReal heightVal );
pfReal getRadius( ) const;
pfReal getHeight( ) const;
// Evaluators
void evalPt(
pfReal
void evalNorm( pfReal
};
u, pfReal v, pfRVec3 &pnt );
u, pfReal v, pfRVec3 &norm );
Main Features of the Methods in pfCylinderSurface
pfCylinderSurface( radius, height ) constructs a cylinder with the specified height and
radius. By default, the z axis is the cylinder’s axis and the cylinder is centered on the
origin, extending in the positive and negative z directions for one-half the height.
For the default orientation, u measures the angle from the x-z plane in a counterclockwise
direction as you look down on the x-y plane and v measures the distance along the z-axis.
The default radius is 1 and the default height is 2.
390
007-1680-100
Parametric Surfaces
pfTorusSurface
The pfTorusSurface class provides methods to describe a torus. Figure 9-13 illustrates a
torus, and how it is parameterized in pfTorusSurface.
A torus can be defined geometrically as the surface in space that is swept by moving a
circle, the minor circle, through space such that its center lies on a second circle, the major
circle, and the planes of the two circles are always perpendicular to each other, with the
plane of the minor circle aligned along radii of the major circle. The parametrization of
the surface is that u represents a position on the major circle and v represents a position
on the minor circle.
z
Origin
pnt
u
y
v
x
Major radius
Figure 9-13
Minor radius
Torus Parameterization
Class Declaration for pfTorusSurface
The class has the following main methods:
class pfTorusSurface : public pfParaSurface
{
public:
// Creating and destroying
007-1680-100
391
9: Higher-Order Geometric Primitives
pfTorusSurface( );
pfTorusSurface( pfReal majorRadius, pfReal minorRadius );
virtual ~pfTorusSurface();
// Accessor functions
void setMajorRadius( pfReal majorRadiusVal );
void setMinorRadius( pfReal minorRadiusVal );
pfReal getMajorRadius( ) const;
pfReal getMinorRadius( ) const;
// Evaluators
virtual void evalPt(
pfReal
virtual void evalNorm( pfReal
}
u, pfReal v, pfRVec3 &pnt );
u, pfReal v, pfRVec3 &norm );
Main Features of the Methods in pfTorusSurface
The constructor pfTorusSurface( majorRadius, minorRadius ) defines a torus with the
specified radii such that the major circle is in the x-y plane and the minor circle is initially
in the x-z plane. The default value for the major radius is 1; the default for the minor
radius is 0.1.
392
007-1680-100
Parametric Surfaces
pfConeSurface
You can define a cone geometrically by sweeping a circle along an axis in a way similar
to the way a cylinder is defined; however, as the circle is swept along the axis, the radius
changes linearly with distance.
The parameterization of a point on an pfConeSurface is that u measures the angle, in
radians, of the point on the circle, and that v measures the distance along the axis from
the origin. To truncate a cone, yielding a frustum, adjust the value for v.
z
y
pnt
Origin
v
Height
u
Half-height
Radius
Figure 9-14
007-1680-100
x
Cone Parameterization
393
9: Higher-Order Geometric Primitives
Class Declaration for pfConeSurface
The class has the following main methods:
class pfConeSurface : public pfParaSurface
{
public:
// Creating and destroying
pfConeSurface( void );
pfConeSurface( pfReal radius, pfReal height );
virtual ~pfConeSurface();
// Accessor functions
void setRadius( pfReal radius ) ;
void setHeight( pfReal height );
pfReal getRadius( ) const;
pfReal getHeight( ) const;
// Evaluators
void evalPt(
pfReal
void evalNorm( pfReal
}
u, pfReal v, pfRVec3 &pnt );
u, pfReal v, pfRVec3 &norm );
Main Features of the Methods in pfConeSurface
The constructor pfConeSurface( radius, height ) creates a parametric cone with the
specified height and a circular base with the specified radius. By default, the base of the
cone is parallel to the x-y plane and centered on the z axis and the apex of the cone is on
the positive z-axis. The cone extends from the origin in the positive and negative z
directions for one half the height. The default for the radius of the base is 1 and the
default height is 2.
394
007-1680-100
Parametric Surfaces
Swept Surfaces
The class pfSweptSurface provides methods to describe a general swept surface. Three
examples of swept surfaces have been presented: a cylinder, a torus, and a cone. In the
first two cases a simple cross-section, a circle of constant radius, was swept along a path.
For a cone, the radius of the circle varied according to a simple profile.
To describe a swept surface, you specify a path, a cross section, and a coordinate frame in
which the graph of the cross section is drawn at each point on the path. The
parameterization of the surface is that u denotes the position along the path and v
denotes the position on the cross-section curve. You can also specify a profile, which
adjusts the size of the cross-section curve. Thus, for example, with a simple profile
method you could generate a sphere from a straight-line path and a circular cross section.
Figure 9-15 illustrates the feature of a swept surface.
007-1680-100
395
9: Higher-Order Geometric Primitives
Cross section
y
x
b
P at
h
t
Figure 9-15
396
Swept Surface: Moving Reference Frame and Effect of Profile Function
007-1680-100
Parametric Surfaces
Orientation of the Cross Section
Unlike the examples of the cylinder, torus, and cone, the cross-section in an
pfSweptSurface generally is not necessarily perpendicular to the path. You set the
orientation of the cross-section with two additional instances of pfCurve3d. For a point
on the path corresponding the parameter value t0, the vectors on these two additional
curves that have the same parameter value define the local coordinate system used to
draw the profile: one vector defines the normal to the plane of the graph, the second the
x axis for the graph, and their cross product determines the direction of the y axis for the
graph. For more details, see the discussion of the constructor below.
Class Declaration for pfSweptSurface
The class has the following main methods:
class pfSweptSurface : public pfParaSurface
{
public:
// Creating and destroying
pfSweptSurface( void );
pfSweptSurface( pfCurve3d *crossSection,
pfCurve3d *_path,
pfCurve3d *_t,
pfCurve3d *_b,
pfScalar *_profile );
virtual ~pfSweptSurface( );
// Accessor functions
void setCrossSection( pfCurve3d *_crossSection );
void setPath( pfCurve3d *_path );
void setT( pfCurve3d *_tng );
void setB( pfCurve3d *_b );
void setProf( pfScalar *_profile );
pfCurve3d *getCrossSection() const;
pfCurve3d *getPath() const;
pfCurve3d *getT() const;
pfCurve3d *getB() const;
pfScalar *getProf()const;
virtual void evalPt( pfReal u, pfReal v, pfRVec3 &pnt );
};
007-1680-100
397
9: Higher-Order Geometric Primitives
Main Features of the Methods in pfSweptSurface
pfSweptSurface( crossSection, path, t, b, profile )
Defines a swept surface with the given path, cross section, and profile.
The arguments t and b are vector-valued functions of the path’s
parameter. They define the orientation of the profile at each point on the
path.
The orientation at a particular point on the curve is determined by
rendering the graph of crossSection in the coordinate plane
perpendicular to t, which locally defines the z axis of an x-y-z
coordinate system. The x axis is defined by the projection of b onto the
plane, and the y axis forms a right-hand coordinate system with the
other two axes. The cross section is plotted in the x-y plane.
If you specify a NULL value for profile, crossSection does not vary along
path.
evalPt( u, v, pnt )
Calculates the point on the surface, pnt, as the vector sum of (a) the point
on the path corresponding to the value u and (b) the point on the cross
section corresponding to the value v. The vector locating the point on the
cross section is scaled by the value at u of the profile function, if profile is
not NULL.
398
007-1680-100
Parametric Surfaces
pfFrenetSweptSurface
As a convenience, the class pfFrenetSweptSurface allows you to use the Frenet frame of
the path to define the orientation vectors in a swept surface. The Frenet frame is defined
by the three unit vectors derived from the tangent, the principal normal, and their cross
product. This set of vectors facilitates orienting the cross section perpendicularly to the
path at every point.
Note: The path for an pfFrenetSweptSurface must be at least a cubic to allow for the
principal normal calculation, which requires a second derivative.
Class Declaration for pfFrenetSweptSurface
The class has the following main methods:
class pfFrenetSweptSurface : public pfSweptSurface
{
public:
// Accessor functions
pfFrenetSweptSurface( void );
pfFrenetSweptSurface( pfCurve3d *crossSection,
pfCurve3d *path,
pfScalar *profile );
virtual ~pfFrenetSweptSurface( );
// Accessor functions
void set( pfCurve3d *crossSection,
pfCurve3d *path,
pfScalar *profile );
};
Main Features of the Methods in pfFrenetSweptSurface
The arguments of the constructor for pfFrenetSweptSurface are the same as for
pfSweptSurface and have the same effects, except for the orientation vectors, which are
set to be the tangent and principal normal to path, and so do not appear as arguments.
Use the inherited method evalPt() to locate points on the surface.
007-1680-100
399
9: Higher-Order Geometric Primitives
Making a Modulated Torus With pfFrenetSweptSurface
The following code uses an pfFrenetSweptSurface to define a torus whose minor radius
varies with position on the ring. Other instances of pfFrenetSweptSurface appear in
repTest.
// Scalar curve used by the swept surface primitive
static pfReal profile( pfReal t )
{
return 0.5*cos(t*5.0) + 1.25;
};
pfCircle3d
*cross =
new pfCircle3d( 0.75, new pfRVec3( 0.0, 0.0, 0.0)
);
pfCircle3d
*path
=
new pfCircle3d( 1.75, new pfRVec3( 0.0, 0.0, 0.0)
);
pfFrenetSweptSurface *fswept =
new pfFrenetSweptSurface( cross, path, profile );
fswept->setHandednessHint( TRUE );
Ruled Surfaces
A ruled surface is generated from two curves in space, both parameterized by the same
variable, u. A particular value of u specifies a point on both curves. A ruled surface is
defined by connecting the two points with a straight line parameterized by v. The
parameterization of the resulting surface is always the unit square in the UV plane,
regardless of the parameterizations of the original curves.
400
007-1680-100
Parametric Surfaces
c2(u)
c1(u)
(1-v)c1(u) + v c2(u)
Figure 9-16
Ruled Surface Parameterization
A bilinear interpolation of four points is perhaps the simplest example of a ruled surface,
one for which the “curves” that define the surface are in fact straight lines. Thus, you
connect two pairs of points in space with lines and then develop the ruled surface. For a
bilinear interpolation, the parameterization by u and v is such that, if one of them is held
constant, a point “moves” along the connecting straight line at a uniform speed as the
other parameter is varied.
Class Declaration for pfRuledSurface
The class has the following main methods:
class pfRuledSurface : public pfParaSurface
{
public:
// Creating and destroying
pfRuledSurface();
pfRuledSurface( pfCurve3d *c1, pfCurve3d *c2 );
virtual ~pfRuledSurface();
// Accessor functions
void setCurve1( pfCurve3d *_c1 );
void setCurve2( pfCurve3d *_c2 );
pfCurve3d *getCurve1( ) const;
pfCurve3d *getCurve2( ) const;
// Evaluators
void evalPt(
};
007-1680-100
pfReal
u, pfReal v, pfRVec3 &pnt );
401
9: Higher-Order Geometric Primitives
The constructor pfRuledSurface( c1, c2 ) creates an instance of a ruled surface defined by
the two curves c1 and c2.
Coons Patches
A Coons patch is arguably the simplest surface you can define from four curves whose
endpoints match and form a closed loop. Think of the four curves as defining the four
sides of the patch, with one pair on opposite sides of the patch defining the top and
bottom curves and the other pair defining the left and right curves (see Figure 9-17). The
top and bottom curves are parameterized by u, and the left and right curves by v. Thus,
u is the “horizontal” coordinate and v the “vertical” coordinate.
The patch is made by
1.
Adding the points on the ruled surface defined by the top and bottom curves to the
points on the ruled surface defined by the left and right curves.
2. Subtracting the bilinear interpolation of the four corner points.
Figure 9-17 illustrates the construction. To understand the result, notice that, after you
add the two ruled surfaces, each side of the boundary of the resulting surface is the sum
of the original bounding curve and the straight line connecting the bounding curve’s
endpoints. The straight line was introduced by the construction of the ruled surface that
did not include the boundary curve. Subtracting the bilinear interpolation eliminates the
straight-line components of the sum, leaving just the original four curves as the
boundary of the resulting surface.
402
007-1680-100
Parametric Surfaces
Top & bottom curves
Left & right curves
z
z
z4
z4
z3
z1
z3
z1
y
z2
y
z2
x
x
Ruled surfaces
z
z
z4
z4
z3
z1
z3
z1
y
z2
y
x
z2
x
2z4
Sum
2z3
z
2z1
z4
Subtract a bilinear
interpolation from the sum
of the ruled surfaces
2z2
z3
z1
y
z2
Bilinear
interpolation
z
z4
x
z3
z1
y
z2
Coons patch,bounded
by left, right, top &
bottom curves
x
Figure 9-17
007-1680-100
Coons Patch Construction
403
9: Higher-Order Geometric Primitives
Class Declaration for pfCoonsSurface
The class has the following main methods:
class pfCoonsSurface : public pfParaSurface
{
public:
pfCoonsSurface( );
pfCoonsSurface( pfCurve3d *right,
pfCurve3d *left,
pfCurve3d *bottom, pfCurve3d *top );
virtual ~pfCoonsSurface( );
// Accessor functions
void setRight( pfCurve3d *right );
void setLeft( pfCurve3d *left );
void setBottom( pfCurve3d *bottom );
void setTop( pfCurve3d *top );
pfCurve3d*
pfCurve3d*
pfCurve3d*
pfCurve3d*
getTop() const;
getBottom() const;
getLeft() const;
getRight() const;
// Surface point evaluator
void evalPt( pfReal u, pfReal v, pfRVec3 &pnt );
};
The constructor pfCoonsSurface( right, left, bottom, top ) creates an instance of a Coons
patch defined by the four curves right, left, bottom, and top. The top and bottom curves are
parameterized by u and the left and right curves are parameterized by v.
NURBS Surfaces
Just as a NURBS curve consists of Bezier curves, a NURBS surface consists of Bezier
surfaces. The set of control parameters is essentially the same for the curves and surfaces:
a set of knots, a control hull, and a set of weights. However, for a NURBS surface, the
knots form a grid in the coordinate system of the surface; that is, in the u-v plane, and the
control hull is a grid of points in space that loosely defines the surface.
Understanding a Bezier surface helps you understand and use a NURBS surface. A
Bezier surface is defined essentially as the surface formed by sweeping a Bezier cross
404
007-1680-100
Parametric Surfaces
section curve through space, along a path defined by a Bezier curve. But, unlike an
pfSweptSurface, the shape of the cross-section can be changed.
You define a Bezier surface as follows:
1.
Start with a Bezier curve in space: the cross section parameterized by u.
2. Define a family of Bezier curves, a set of paths all of which are parameterized by v,
that start at the control points of the initial cross section.
For each value of v, the set of control points defines a Bezier curve. As v changes, the
cross-sectional curve “moves” through space, changing shape and defining a Bezier
surface.
A NURBS surface joins Bezier surfaces in a smooth way, similar to NURBS curves joining
Bezier curves. The class pfNurbSurface provides methods to describe a NURBS surface.
Class Declaration for pfNurbSurface
The class has the following main methods:
class pfNurbSurface : public pfParaSurface
{
public:
// Creating and destroying
pfNurbSurface( void );
virtual ~pfNurbSurface( );
// Accessor functions
void setControlHull( int iu, int iv, const pfRVec3 &p );
void setControlHull( int iu, int iv, const pfRVec4 &p );
void setWeight( int iu, int iv, pfReal w );
void setUknot( int iu, pfReal u );
void setVknot( int iv, pfReal v );
void setControlHullUSize( int s );
void setControlHullVSize( int s );
// Get the same parameters
pfRVec3& getControlHull( int iu, int iv) ;
int
getControlHullUSize( void );
int
getControlHullVSize( void );
pfReal getWeight( int iu, int iv)
pfReal getUknot( int iu);
pfReal getVknot( int iv);
int
getUknotCount( void );
007-1680-100
405
9: Higher-Order Geometric Primitives
int
int
int
void
void
void
void
getVknotCount( void );
getUorder( void ) ;
getVorder( void ) ;
removeControlHullElm(int ui, int iv);
removeUknot(int iu);
removeVknow(int iv);
flipUV();
// Evaluator
virtual void
virtual void
virtual void
virtual void
};
evalPt(
evalDu(
evalDv(
evalNorm(
pfReal
pfReal
pfReal
pfReal
u,
u,
u,
u,
pfReal
pfReal
pfReal
pfReal
v,
v,
v,
v,
pfRVec3
pfRVec3
pfRVec3
pfRVec3
&pnt );
&Du );
&Du );
&norm );
Main Features of the Methods in pfNurbSurface
The member functions are essentially the same as those for pfNurbCurve3d (see
“NURBS Curves in Space” on page 373), however:
•
The hull is a grid of pfRVec3s indexed by i and j.
•
The set of knots is defined by points on the u and v axes.
•
There are B-spline basis functions (of possibly differing orders) associated with each
coordinate direction.
Note: pfNurbSurface redefines the virtual evaluators inherited from pfParaSurface for
tangent and normal vectors; the methods use the NURBS equation rather than finite,
central differences.
Indexing Knot Points and the Control Hull
Indexing of knot points in coordinate space and control hull points in three-dimensional
space is illustrated in Figure 9-18. The indexing works as for gluNurbsSurface, that is, as
follows:
406
•
iu indexes knots on the u axis. The correspondence is established by setUknot().
•
iv indexes knots on the v axis.The correspondence is established by setVknot().
•
Each (iu,iv) thus indexes a knot point in the u-v plane.
007-1680-100
Parametric Surfaces
setControlHull (iu, iv, p)
z
(3,2)
(0,0)
y
x
v
setUknot (iu, u)
setVknot ( iv, v)
v4
v3
iv = 0,1,2,3...
(3,2)
v2
v1
v0
(0,0)
u
u0 u 1
u2
u3
u4 u5
u6
iu = 0,1,2,3...
•
Each (iu,iv) also indexes a point on the control hull in three-dimensional space. The
correspondence is established by setControlHull().
•
Thus, setUknot(), setVknot(), and setControlHull() establish a correspondence
between an index pair (iu,iv) a knot point (uiu viv), and a point on the control hull in
three-dimensional space.
Figure 9-18
007-1680-100
NURBS Surface Control Hull Parameterization
407
9: Higher-Order Geometric Primitives
Equation Used to Calculate a NURBS Surface
Indexing is determined by the following equation that OpenGL Performer uses to
calculate a NURBS surface (the index i corresponds to iu in the API, and j corresponds to
iv):
∑ B (u)B (v)C
p(u, v) = ---------------------------------------------∑ B (u)B (v)W
m
i
n
j
ij
m
i
n
j
ij
i, j
i, j
where
•
p(u, v) is a point on the surface
•
B im(u) is the ith B-spline basis polynomial of degree m
•
C ij is a control point
•
W ij is the weight for the control point
Alternative Equation for a NURBS Surface
A NURBS surface can also be developed from the following alternative expression:
∑ B (u)B (v)W C
p(u, v) = ----------------------------------------------------∑ B (u)B (v)W
m
i
n
j
ij
ij
i, j
m
i
n
j
ij
i, j
For this case, you must change the coordinates of the control points to get the same
surface from OpenGL Performer. You convert the coordinates of the control points from
(x,y,z,w) to (wx,wy,wz,w).
408
007-1680-100
Parametric Surfaces
Sample of a Trimmed pfNurbSurface From repTest
The following code fragment form the repTest sample application illustrates an instance
of an pfNurbSurface. Toward the end of the example, an optional pfNurbCurve2d trim
curve is created.
int i, j;
pfNurbSurface *nurb = new pfNurbSurface;
// Control hull dimensions
#define USIZE 4
#define VSIZE 5
// Set up the control hull size because we know a priori how big
// the nurb is. The next two lines are used for space
// efficiency but are functionally unnecessary.
nurb->setControlHullUSize(USIZE);
nurb->setControlHullVSize(VSIZE);
// Make the control hull be an oscillating grid
for ( i = 0; i < VSIZE; i++ )
{
pfReal y = i/(float)(VSIZE - 1) * 2*M_PI - M_PI;
for ( j = 0; j < USIZE; j++ )
{
pfReal x = j/(float)(USIZE - 1) * 2*M_PI - M_PI;
pfReal val = 6*pow( cos(sqrt(x*x + y*y)), 2.0);
// Make the control hull a box, j maps to u and i maps to v
nurb->setControlHull( i, j, pfRVec3( x, y, val));
// Add the weights
nurb->setWeight( i, j, 1.0 );
}
}
// Add the knot
nurb->setUknot(
nurb->setUknot(
nurb->setUknot(
nurb->setUknot(
nurb->setUknot(
nurb->setUknot(
007-1680-100
points
0, 0.0
1, 0.0
2, 0.0
3, 0.0
4, 1.0
5, 1.0
);
);
);
);
);
);
409
9: Higher-Order Geometric Primitives
nurb->setUknot( 6,
nurb->setUknot( 7,
1.0 );
1.0 );
nurb->setVknot(
nurb->setVknot(
nurb->setVknot(
nurb->setVknot(
nurb->setVknot(
nurb->setVknot(
nurb->setVknot(
nurb->setVknot(
0.0
0.0
0.0
0.0
1.0
1.0
1.0
1.0
0,
1,
2,
3,
4,
5,
6,
7,
);
);
);
);
);
);
);
);
// Only trim reps in the first row
if ( nVersions <= 0 )
{
// Add a super quadric trim curve
pfSuperQuadCurve2d *trimCircle0 = new pfSuperQuadCurve2d( 0.25, new
pfRVec2(0.25, 0.50), 2.0 );
nurb->addTrimCurve( 0, trimCircle0, NULL );
// make a 4-th order nurb trim curve
pfNurbCurve2d *l = new pfNurbCurve2d;
l->setKnot(0,0.0);
l->setKnot(1,0.0);
l->setKnot(2,0.0);
l->setKnot(3,0.0);
l->setKnot(4,1.0);
l->setKnot(5,1.0);
l->setKnot(6,1.0);
l->setKnot(7,1.0);
l->setControlHull(0,pfRVec2(0.50,0.50));
l->setControlHull(1,pfRVec2(0.90,0.10));
l->setControlHull(2,pfRVec2(0.90,0.90));
l->setControlHull(3,pfRVec2(0.50,0.50));
nurb->addTrimCurve( 1, l, NULL
}
410
);
007-1680-100
Parametric Surfaces
Hermite-Spline Surfaces
Hermite-spline surfaces interpolate a grid of points; that is, they pass through the set of
specified points under the constraint that you supply the tangents at each point in the u
and v directions and the mixed partial derivative at each point. This surface definition is
the natural generalization of Hermite-spline curves, discussed in “Hermite-Spline
Curves in the Plane” on page 359.
tuv
tv
tuv
tv
tu
tuv
tuv
tv
tv
tu
tuv
tu
tv
tv
tuv
tu
tv
tuv
tv
tu
tu
tuv
tu
tuv
tv
tu
tu
Figure 9-19
Hermite Spline Surface With Derivatives Specified at Knot Points
Hermite-spline surfaces are made of Hermite patches (see Figure 9-19). A bicubic Hermite
patch expands the definition of a bilinear interpolation to include specification of first
derivatives and mixed partial derivatives of the surface at each of the four corners. The
adjective “bicubic” in the name of the patches refers to the mathematical definition,
which includes products of the cubic Hermite polynomials that define a Hermite-spline
curve.
An advantage of including the derivatives to constrain the surface is that it is simple to
combine the patches into a smooth composite surface, that is, into a Hermite-spline surface.
007-1680-100
411
9: Higher-Order Geometric Primitives
Class Declaration for pfHsplineSurface
The class has the following main methods:
class pfHsplineSurface : public pfParaSurface
{
public:
// Creating and destroying
pfHsplineSurface();
pfHsplineSurface( pfReal *_p,
pfReal *_tu, pfReal *_tv, pfReal *_tuv,
pfReal *_uu, pfReal *_vv,
int uKnotCount, int vKnotCount );
virtual ~pfHsplineSurface();
// Accessor functions
pfRVec3& getP( int i, int j );
pfRVec3& getTu( int i, int j );
pfRVec3& getTv( int i, int j );
pfRVec3& getTuv( int i, int j );
pfReal getUknot( int i );
pfReal getVknot( int j );
int
getUknotCount();
int
getVknotCount();
pfBool getCylindrical();
void setAll( pfReal *p,
pfReal *tu,
pfReal *tv,
pfReal *tuv,
pfReal *uu,
pfReal *vv,
int uKnotCount,
int vKnotCount );
void setCylindrical( pfBool cylinderical );
// Surface point evaluator
void evalPt( pfReal u, pfReal v, pfRVec3 &pnt );
};
Main Features of the Methods in pfHsplineSurface
The pfHsplineSurface class has two important methods, the constructor and
set/getCylinderical().
412
007-1680-100
Meshes
The pfHsplineSurface constructor has the following arguments:
_p
Specifies the grid of points on the surface.
_tu, _tv, and _tuv
Specify, respectively, the corresponding tangents in the u and v
directions and the mixed partials.
The indexing of each of the arrays _p, _tu, _tv, and _tuv is as follows: the
x, y, and z components of each vector are grouped in that order, and the
sequence of points is defined so that the vKnotCount index changes
more rapidly.
uKnotCount and vKnotCount
Specify the number of points in the grid. The surface is made of
(uKnotCount-1) × (vKnotCount-1) Hermite patches.
_uu and _vv
Define the knot points, the parameter values corresponding to the patch
corners; thus, they have uKnotCount and vKnotCount elements,
respectively.
setCylindrical() and getCylindrical()
Control the flag for whether the coordinates and derivatives are
assumed to be in cylindrical coordinates.
Meshes
A mesh, encapsulated in OpenGL Performer by the high-level class pfMesh, is used to
store information about the topology of an object. A pfMesh is derived from a pfObject
for memory allocation purposes. A pfMesh stores the object as a collection of vertices and
flat faces (a vertex being encapsulated as a pfMeshVertex and a face as a pfMeshFace). It
is possible to query the mesh to determine the information about neighbors of a given
vertex or a face. This information is required by various algorithms—for example, for
evaluating subdivision surfaces or for simplification of objects.
You can build a pfMesh for a given object in several ways:
007-1680-100
•
You can use the function pfdAddNodeToMesh() that parses a given node and adds
all its geometry to the mesh.
•
You can write your own traverser and for each pfGeoSet you can call the function
pfMeshAddGeoSet().
413
9: Higher-Order Geometric Primitives
•
You can call the function pfMeshAddTriangle() or pfMeshAddFace() and add each
face separately.
•
Where you would like to have full control over the topology stored in the pfMesh,
you can directly set the array of faces and vertices. To do that, you can create the
corresponding arrays by calling pfMeshNumFaces() and pfMeshNumVertices()
and then set each face and vertex separately. You can get the pointer to the array
item by calling the function pfGetMeshVertex() and pfGetMeshFace(),
respectively.
When using the function pfMeshAddGeoSet(), you can specify the pfGeoSet, the
number of a mesh part, and the current transformation. The information about parts can
be used to mark edges. Each edge is automatically marked with the following flags:
Flag
Description
PFM_EDGE_NORMAL
An edge with two adjacent faces
+PFM_EDGE_BOUNDARY
An edge with one adjacent face to the left of the edge
–PFM_EDGE_BOUNDARY
An edge with one adjacent face to the right of the
edge
PFM_EDGE_SPLIT
An edge with more than one adjacent face (in the case
of non-manifold surfaces).
You can mark edges between different parts as PFM_EDGE_CREASE if the flag
PFM_FLAG_CREASES_BETWEEN_PARTS, described later in this section, is set. Note that
edges between faces with different pfGeoStates are also usually marked as creases.
The part index and the transformation is also provided to functions
pfMeshAddTriangle() and pfMeshAddFace(). In addition, you also need to set the
pfGeoState of the triangle or face.
When setting vertices and faces directly and accessing them through a pointer, you need
to set all vertex and face parameters. See the later sections “Mesh Faces” on page 416 and
“Mesh Vertices” on page 417.
You must call the function pfMeshSplitVertices() after all vertices are set, regardless of
the method used to build the mesh. This function performs extra post-processing that
cleans the mesh in cases where the object is not a manifold (see section “Mesh Vertices”
on page 417 for more details).
414
007-1680-100
Meshes
You can set the following flags to 0 or 1:
007-1680-100
Flag
Description
PFM_FLAG_TRIANGULATE
If this flag is set, quads and polygons
processed by pfMeshAddGeoSet()
are triangulated. This flag can be
used for Loop subdivision. It is off by
default.
PFM_FLAG_TEST_ORIENTATION
If this flag is set, the normal at the
first vertex of a face is compared with
the face normal. If they point in the
opposite directions, the order of
vertices is is changed. This flag is off
by default.
PFM_FLAG_CREASES_BETWEEN_PARTS
If this flag is set, it automatically
marks edges between different parts
as creases. It is on by default.
PFM_FLAG_CREASES_BETWEEN_GEOSTATES
If this flag is set, it automatically
marks edges between pfGeoSets
with different pfGeoStates. It is on by
default.
PFM_FLAG_QUAD_TSTRIPS
If this flag is set, triangle strips
processed by pfMeshAddGeoSet()
are converted to quads even if the
quads are not planar. It is off by
default.
PFM_FLAG_USE_VERTEX_GRID
If this flag is set, a grid is used when
adding faces and checking for
existing vertices. You must set the
bounding sphere of the mesh to use
the grid established by
pfMeshGridBsphere(). It is off by
default.
415
9: Higher-Order Geometric Primitives
You can set the following values:
Value
Description
PFM_EPSILON
Sets the epsilon to be used when comparing vertices.
The default is 1e-6 or if the grid is defined, 1e-6
multiplied by a ratio of the grid diameter to the grid
resolution along X axis.
PFM_VERTEX_GRID_SIZE_X
See the description for flag
PFM_VERTEX_GRID_SIZE.
PFM_VERTEX_GRID_SIZE_Y
See the description for flag
PFM_VERTEX_GRID_SIZE.
PFM_VERTEX_GRID_SIZE_Z
See the description for flag
PFM_VERTEX_GRID_SIZE.
PFM_VERTEX_GRID_SIZE
Sets the resolution of the grid used for comparing
vertices. You can set the grid size one value at the time
or all three coordinates at once. The default grid size
is 32x32x32.
You can also query the bounding box of the mesh using the function pfGetMeshBbox().
This bounding box is computed automatically around the mesh vertices.
You can use the function pfMeshUpdateMesh() to automatically detect which vertices
have been changed and to mark them and the corresponding faces. This can be used in
case of dynamically controlled meshes (see the section “Subdivision Surfaces” on
page 420 for an example). Since each vertex stores both the pointer to the original vertex
location and the vertex coordinates, any changes can be easily detected by comparing
these values.
Mesh Faces
OpenGL Performer uses mesh faces and its vertices to describe a mesh, which in turn
describes the topology of an object. The high-level class pfMeshFace encapulsates a face.
Each face contains the following:
•
416
An array of face vertices
The class pfMeshVertex represents a vertex. You can use the functions
pfMeshFaceNumVerts() and pfMeshFaceVertex() to set face vertices. The
parameters of the function pfMeshFaceVertex() are the vertex index in the array
007-1680-100
Meshes
stored with the face (which vertex you want to set) and the vertex index in the array
of vertices associated with a pfMesh. The latter index identifies the vertex. You can
use the functions pfGetMeshFaceNumVerts() and pfGetMeshFaceVertex() to
query face vertices.
•
A set of texture coordinates
You can use the function pfMeshFaceTexCoord() to specify texture coordinates as
vertices.
•
A pfGeoSet
•
A part index
The part index is used by the pfMesh to determine which edges are marked as
smooth and which, as creases (see the preceding section “Meshes” on page 413)
•
Flags
There is only one flag used by a pfMeshFace: PFMF_FLAG_FACE_CHANGED. You set
this flag with the function pfMeshUpdateMesh() to indicate that a vertex of the face
has changed position.
Mesh Vertices
A pfMeshVertex is a high-level OpenGL Performer class used in a pfMesh to store
information about vertices. A pfMeshVertex stores the following:
•
Vertex coordinates
•
The pointer to the original vertex coordinates (to be able to detect local changes in
the mesh)
•
An array of vertex neighbors
•
A set of binary flags
•
An index of another vertex at the same location (used for non-manifolds,as
described later in the subsection “Vertex Neighbors” on page 418)
This section describes these items in the following subsections:
007-1680-100
•
“Vertex Coordinates”
•
“Vertex Neighbors”
•
“Binary Flags”
417
9: Higher-Order Geometric Primitives
Vertex Coordinates
You can use the functions pfMeshVertexCoord() and pfMeshVertexCoordPtr() to set the
vertex coordinates and the pointer to the original coordinates. The value set by
pfMeshVertexCoord() should correspond to the value at the location set by
pfMeshVertexCoordPtr(). If the specified pointer is NULL (for example, where a
pfGeoSet was under a pfSCS node), it is not possible to automatically detect changes in
position of vertices and you must modify all vertices manually by calling the function
pfMeshVertexCoord(). If you do not plan to animate the mesh, you can set the pointer
to NULL.
Vertex Neighbors
Each vertex stores an array of its neighbors. You can query the neighbors using the
functions pfMeshVertexNumNeighbors() and pfMeshVertexNeighbor(). The structure
of pfMeshVertexNeighbor() is defined in pf.h as follows:
typedef struct
{
int vertex;
int face;
short int edgeType;
short int next, prev;
} pfMeshVertexNeighbor;
This structure consists of the vertex index (in the array of vertices in pfMesh), face index
(in the array of faces in pfMesh), edge type, and index of the next and previous neighbor
in the array of neighbors.
418
007-1680-100
Meshes
The edge can be one of the following types:
Edge Type
Description
PFM_EDGE_NORMAL
Specifies a smooth edge with two adjacent faces.
PFM_EDGE_CREASE
Specifies a sharp edge with two adjacent faces.
+PFM_EDGE_BOUNDARY
Specifies an edge with one adjacent face to the left of the
edge.
–PFM_EDGE_BOUNDARY
Specifies an edge with one adjacent face to the right of the
edge.
The pertinent face is the face to the left of the edge (unless the edge is marked as
–PFM_EDGE_BOUNDARY). The next neighbor is part of this face and the previous
neighbor is part of the face to the right. If the edge is of type –PFM_EDGE_BOUNDARY, the
next neighbor points to the corresponding edge of type +PFM_EDGE_BOUNDARY and the
previous neighbors of the edge of type +PFM_EDGE_BOUNDARY point to the the
corresponding edge of type +PFM_EDGE_BOUNDARY. Note that in the case of manifolds
each vertex has exactly zero or two boundary edges.
In the case of arbitrary surfaces, there may be more than one loop of neighbors (if you
follow the next links). That is why the class pfMesh provides the function
pfMeshSplitVertices(), which splits each vertex where the surface behaves as
non-manifold. The vertex is split into several vertices, each having a single loop of
ordered neighbors and the same position.
You can access the next and previous neighbor for a given neighbor with vertex indexed
v1 by calling the functions pfPreviousNeighborMeshVertex() and
pfNextNeighborMeshVertex(). You can also determine the index of such a neighbor in
the array of neighbors by calling pfPreviousNeighborIndexMeshVertex() and
pfNextNeighborIndexMeshVertex().
007-1680-100
419
9: Higher-Order Geometric Primitives
Binary Flags
The following flags can be set for each pfMeshVertex:
Flag
Description
PFMV_FLAG_VERTEX_CHANGED
Set by the function
pfMeshUpdateMesh() when the
coordinate stored at the vertex does not
match the value at the coordinate pointer.
PFMV_FLAG_VERTEX_NEIGHBOR_CHANGED Set by the function
pfMeshUpdateMesh() when the
position of any neighbor changes.
PFMV_FLAG_VERTEX_FACE_CHANGED
Set by the function
pfMeshUpdateMesh() when the
position of any vertex on any of the faces
associated with this vertex changes.
PFMV_FLAG_VERTEX_SPLIT
Set by the function
pfMeshSplitVertices() when the surface
around the vertex is not manifold and the
vertex is split into several vertices (see
the preceding section “Vertex
Neighbors” on page 418).
Subdivision Surfaces
A subdivision surface is specified by a control mesh consisting of a set of connected faces
(usually triangles or quads). The control mesh is stored as a pfMesh. The OpenGL
Performer class pfSubdivSurface is used to define a subdivision surface.
Before rendering a subdivision surface, each face is recursively subdivided into a finer
mesh. For example, in the case of Loop subdivision, each triangle is subdivided into four
new triangles at each subdivision step. The positions of both original vertices and newly
introduced vertices are determined by the subdivision rules. Usually the position is a
linear combination of the positions of the original vertices of the face and the vertices of
neighboring faces. The parameters of the linear combination are set in a such a way that,
after a few subdivision steps, the resulting mesh smooths sharp edges. For example, if
the control mesh is a tetrahedron after a few subdivision steps, you get an oval shape.
Optionally, it is possible to mark selected edges not to be smooth. An excellent source of
420
007-1680-100
Subdivision Surfaces
information on subdivision surfaces is the SIGGRAPH 2000 course "Subdivision for
Modeling and Animation" (http://mrl.nyu.edu/publications/subdiv-course2000/).
The remainder of this section describes the following topics:
•
“Creating a Subdivision Surface”
•
“Loop and Catmull-Clark Subdivisions”
•
“Dynamic Modification of Vertices”
•
“The libpfsubdiv Pseudo Loader”
•
“Special Notes”
Creating a Subdivision Surface
The function pfSubdivSurface() creates and returns a handle to a pfSubdivSurface. Like
other pfNodes, pfSubdivSurfaces are always allocated from shared memory and cannot
be created statically on the stack or in arrays. Use the function pfDelete() rather than the
delete operator to delete pfSubdivSurfaces.
To define a subdivision surface, you must specify its control mesh using the function
pfSubdivSurfaceMesh(). The mesh is stored as a pfMesh. A pfMesh is a data structure
that stores the connectivity between individual faces of the mesh. For each vertex and
face, you can access neighboring vertices and faces, respectively. See the preceding
section “Meshes” on page 413 for more details.
To create a pfMesh, you can either specify its faces and the connectivity one by one or use
the function pfdAddNodeToMesh(). This function takes a pfNode, parses all its
pfGeoSets, splits all primitives into planar faces, and adds those faces into the specified
pfMesh (see the pfdAddNodeToMesh man page).
Before you set the mesh using pfSubdivSurfaceMesh(), you need to set various
parameters of the subdivision because different internal data structures may be needed
for different types of subdivision. You use the function pfSubdivSurfaceVal() to set the
values and the function pfSubdivSurfaceFlags() to set the flags.
007-1680-100
421
9: Higher-Order Geometric Primitives
You can set the following values:
Parameter Value
Description
PFSB_SUBDIVISION_METHOD
Can be either PFSB_CATMULL_CLARK
(Catmull-Clark subdivision, the default) or
PFSB_LOOP (Loop subdivision).
PFSB_SUBDIVISION_LEVEL
Specifies the number of subdivision steps. Usually
two or three steps are quite reasonable. The default is
0.
PFSB_MAX_DATA_HEIGHT_SW
Specifies the limit of data structures used to evaluate
the subdivision surface in software. The default is
unlimited. However, if you are running out of
memory for large models evaluated in software, try a
value like 10,000. This would mean that only 10,000
faces would be evaluated at a time.
PFSB_MAX_PBUFFER_HEIGHT
Specifies the maximum height of a pbuffer (off-screen
memory on a GPU) used for evaluating the surface
directly on the graphics hardware (GPU). The default
is 512. It should be a power of two.
You can set the following flags to 0 or 1:
Flag
Description
PFSB_GPU_SUBDIVISION
Allows direct evaluation of the surface on the graphics
hardware (GPU) if set to 1. The hardware must support
floating point fragment shaders (for example, Onyx4 or
Prism systems). The default is 0.
PFSB_CONTROL_VERTICES_DYNAMICAllows the modification of the vertices of the
control surface (only their position) on the fly. See the later
subsection “Dynamic Modification of Vertices” on
page 424 for more information. The default is 0.
PFSB_USE_GEO_ARRAYS
422
Stores the resulting subdivided mesh (in the case of
software evaluation) in pfGeoArrays. Otherwise, regular
pfGeoSets are used. The default is 1.
007-1680-100
Subdivision Surfaces
The following examples set up a subdivision surface and add it into a scene.
Example 1:
pfSubdivSurface *subdivSurface;
pfMesh
*mesh;
subdSurf = pfNewSubdivSurface(arena);
mesh = pfNewMesh(arena);
if(method == PFSB_LOOP)
pfMeshFlags(mesh, PFM_FLAG_TRIANGULATE, 1); // important!
pfdAddNodeToMesh(root, mesh, 0); // part 0
pfSubdivSurfaceVal(subdSurf, PFSB_SUBDIVISION_LEVEL, subdivLevel);
pfSubdivSurfaceVal(subdSurf, PFSB_SUBDIVISION_METHOD, method);
pfSubdivSurfaceFlags(subdSurf, PFSB_GPU_SUBDIVISION, GPUsubdivision);
Example 2:
pfSubdivSurfaceMesh(subdSurf, mesh);
pfAddChild(scene, subdSurf);
pfSubdivSurface *subdivSurface = new pfSubdivSurface;
pfMesh *mesh = new pfMesh;
if(method == PFSB_LOOP)
mesh->setFlags(PFM_FLAG_TRIANGULATE, 1); // important!
pfdAddNodeToMesh(root, mesh, 0); // part 0
subdivSurface->setVal(PFSB_SUBDIVISION_LEVEL, subdivLevel);
subdivSurface->setVal(PFSB_SUBDIVISION_METHOD, method);
subdivSurface->setFlags(PFSB_GPU_SUBDIVISION, GPUsubdivision);
subdivSurface->setMesh(mesh);
scene->addChild(subdivSurface);
You can find sample code in the file
perf/samples/pguide/libpf/C++/subdivSurface.C. or in the file
%PFROOT%\Src\pguide\libpf\C++\subdivSurface.cxx on Microsoft Windows.
007-1680-100
423
9: Higher-Order Geometric Primitives
Loop and Catmull-Clark Subdivisions
The major difference between Loop subdivision and Catmull-Clark subdivision is in the
type of faces they process. Loop subdivision operates exclusively on triangles. Therefore,
the control mesh has to be triangulated. Since a pfMesh does not know whether it is used
in a pfSubdivSurface and what subdivision method is used, you must set the flag
PFM_FLAG_TRIANGULATE on the pfMesh before specifying the faces (see the sample
code in the preceding section “Creating a Subdivision Surface” on page 421).
Catmull-Clark subdivision, on the other hand, operates on quads. If the control mesh
contains non-quads, the first subdivision step is performed using special rules so that an
arbitrary face is divided into a set of quads. Note that this step cannot be performed on
a GPU. This step introduces some discontinuities in the second derivative of the
curvature. Consequently, the surface may appear more bumpy around the edges of the
original control faces. For this reason, start with a control mesh containing only quads for
optimal results. The advantage of Catmull-Clark subdivision is that it is supported by
many modeling tools, including Maya.
Dynamic Modification of Vertices
If the flag PFSB_CONTROL_VERTICES_DYNAMIC is set, you can modify the position of
vertices of the control mesh to animate the surface. You need to call the function
pfSubdivSurfaceUpdateControlMesh() at each frame to confirm the changes. Note that
you cannot have in the node any tranforms specifying the control mesh.
The libpfsubdiv Pseudo Loader
You can subdivide an arbitrary file loaded into OpenGL Performer by using the
libpfsubdiv pseudo loader. It works similarly as the libpftrans or libpfscale
pseudo loader. Suppose that you want to subdivide truck.pfb in Perfly. You enter the
following command:
perfly truck.pfb.0,2,0.subdiv
The syntax of the libpfsubdiv loader is as follows:
<filename>.<method>,<subdivLevel>,<GPUon>.subdiv
424
007-1680-100
Subdivision Surfaces
The value for method is 0 for Catmull-Clark and 1 for Loop subdivision. The subdivLevel
value is usually around 2 and GPUon is 1 if the evaluation should be done fully on a
GPU.
Special Notes
Note the following limitations and anomalies regarding the use of subdivision surfaces:
007-1680-100
•
Subdivision level 0 may result in incorrect normals.
•
Currently, the GPU subdivision supports only control meshes with vertices that
have less than 11 neighbors.
•
Current drivers on Onyx4 , Prism, and Microsoft Windows platforms do not
support super buffers. Therefore, the GPU evaluation of subdivision surfaces on
those systems is too slow.
425
Chapter 10
10. Creating and Maintaining Surface Topology
Most objects in a large model are made of many parametric surfaces. The OpenGL
Performer classes that describe the connectivity of parametric surfaces—that is, their
topology—allow you to “stitch” surfaces together by defining shared boundary curves,
and to propagate surface contact information.
The main purpose for shared-boundary information is to generate tessellations of
adjacent surfaces that are consistent, that is, no cracks develop between any pair of
rendered surfaces. Tessellations are discrete approximations of surfaces in terms of
renderable geometric primitives, typically triangles (see Chapter 11, “Rendering
Higher-Order Primitives: Tessellators”).
These topics are covered in this chapter:
•
“Overview of Topology Tasks” on page 427
•
“Summary of Scene Graph Topology: pfTopo” on page 428
•
“Collecting Connected Surfaces: pfSolid” on page 437
Overview of Topology Tasks
The topology classes provide definitions of boundary curves shared by adjacent
parametric surfaces. Discrete versions of these curves are used by tessellators to prevent
cracks. A rendered image can have artificial cracks due to the following:
•
Difficulty sampling enough points on the boundary between two surfaces so that
mismatches of the tessellations are imperceptible
•
Finite-precision mismatches between coordinates of ideally identical points, for
example at triple junctions where the edges of three surfaces meet at a point
Propagating surface contact information is useful for other tasks, such as
•
007-1680-100
Maintaining consistent normal vectors for adjacent surfaces
427
10: Creating and Maintaining Surface Topology
•
Deforming a surface and consistently deform an adjacent surface
•
Determining whether an edge of a surface is in fact a shared boundary
•
Creating a mirror image of a compound surface (you can use topological
information to reorient the surface)
Summary of Scene Graph Topology: pfTopo
The class pfTopo holds data that indicates whether, and how, two pfParaSurfaces are in
contact. You can create several pfTopos for a particular scene: for example, one each for
subassemblies. A static member of pfTopo lists all the pfTopos that you create.
pfTopo maintains lists of surfaces and boundaries (pfBoundarys) that are shared by an
arbitrary number of surfaces. Figure 10-1 illustrates how these data structures define
relations between pfParaSurfaces.
When an edge has been tessellated, the associated pfBoundary holds a discrete version
of the curve. This discrete version is needed for consistent tessellations because it
specifies one set of boundary vertices for tessellating all the surfaces that share the
boundary. The role of pfBoundary in determining a consistent tessellation is illustrated
in Figure 10-2.
The classes pfTopo and pfBoundary are examples of b-reps, which identify objects in
terms of their bounding objects. pfBoundary is also winged data structures, a particular
form of b-rep.
428
007-1680-100
Summary of Scene Graph Topology: pfTopo
ar
pfP
p
un
fBo
dary
pfP
a
pfP
p
raS
ge
fEd
urfa
ce
pfE
ara
dg
pfE
pfB
e
fac
Sur
pfE
e
pf
pfB
aSu
e
Edg
rfac
pfE
e
dge
ta
Da ture
c
u
r
st
oun
dary
dge
2 trim curves
specified by
2 edges make
up each of the
trim loops in
the figure
oun
dary
dge
u-v
coo
te
na e
rdi spac
-z
x-y odel e
m pac s
s ace
f
sur
Figure 10-1
007-1680-100
Topological Relations Maintained by Topology Classes
429
10: Creating and Maintaining Surface Topology
p
ra
fPa
Sur
pfE
urf
ace
raS
a
pfP
o
pfB
und
fac
e
dge
ary
e
a r
at tu
D ruc
st
dge
pfE
Trim
curves
te
na
di
or
co
v e
u- ac
sp
z
y- el
x- od e s
m ac ce
sp rfa
su
n nt
e
tio te
et
cr e ta is
is c n s
D rfa ese con n)
su pr e tio
re dg el a
(E ss
te
Figure 10-2
430
Consistently Tessellated Adjacent Surfaces and Related Objects
007-1680-100
Summary of Scene Graph Topology: pfTopo
Building Topology: Computing and Using Connectivity Information
Given a set of pfParaSurfaces in a scene graph, there are several ways to develop a set of
shared vertices to be held in pfBoundarys. The following sections describe the topology
construction strategies (beyond the low-fidelity alternative of ignoring topology):
•
“Building Topology Incrementally: A Single-Traversal Build” on page 431
•
“Building Topology From All Scene Graph Surfaces: A Two-Traversal Build” on
page 432
•
“Building Topology From a List of Surfaces” on page 432
•
“Building Topology “by Hand”: Imported Surfaces” on page 432
•
“Summary of Topology Building Strategies” on page 433
Building Topology Incrementally: A Single-Traversal Build
As each surface is tessellated during a traversal, the tessellator checks for previously
tessellated adjacent surfaces, uses existing vertices when it can, and adds necessary data
to topology data structures.
Although OpenGL Performer’s incremental topology building tools attempt to avoid
cracks, they can, in principle, appear: When a surface is added, a new junction on the
boundary of an existing, tessellated surface may occur and the junction point may not be
in the existing tessellation. The tessellation of the added surface introduces the junction
point, necessarily at a finite distance from the existing tessellation, and a crack appears
between the newly and previously tessellated surfaces.
007-1680-100
431
10: Creating and Maintaining Surface Topology
Building Topology From All Scene Graph Surfaces: A Two-Traversal Build
Topology built with two passes is very clean; unlike a single-pass build, in principle no
cracks due to unforeseen junctions can occur. The added cost of performing a
two-traversal build is slight; it is the recommended way to build topology and perform
tessellations if you want high-quality images. When building topology in two traversals,
the following steps occur:
1.
Connectivity of all surfaces is calculated during a topology building traversal of the
scene graph, before a tessellation traversal.
2. The surfaces in the scene are tessellated during a second traversal.
Building Topology From a List of Surfaces
You can explicitly accumulate a list of surfaces for which to build topology and then
tessellate the surfaces. The result is clean tessellations of the surfaces on the list. Cracks
may appear if an adjacent surface was not included in the list.
Building Topology “by Hand”: Imported Surfaces
If you have a set of surfaces for which you know connectivity, you can explicitly develop
the appropriate topological data structures and develop consistent tessellations.
The presence of cracks will depend on how good your input trim curves are. If three
surfaces meet at a junction point that is not the shared endpoint of trim curves, a crack
may appear.
432
007-1680-100
Summary of Scene Graph Topology: pfTopo
Summary of Topology Building Strategies
Table 10-1 lists the methods required for each of the topology building strategies. See
“Base Class pfTessellateAction” on page 443 for more information about the tessellation
methods listed.
Table 10-1
Topology Building Methods
Topology Building Strategy
Methods
Ignore topology information and let
cracks appear as they will.
1. Do not create an pfTopo or build topology.
2. pfTessellateAction::setBuildTopoWhileTess(FALSE).
3. pfdTessellateGeometry(root, tessAction)
Build topology incrementally.
1. Create an pfTopo.
2. pfTessellateAction::setBuildTopoWhileTess(TRUE).
3. pfTessellateAction::setTopo(topo).
4. pfdTessellateAction(root, tessAction).
Two-traversal build.
1. Create an pfTopo.
2. pfTopo::buildTopologyTraverse(root).
3. pfTessellateAction::setBuildTopoWhileTess(FALSE).
4. pfdTessellateAction(root, tessAction).
Assemble a list of surfaces, build the
topology, and then tessellate.
1. Create an pfTopo.
2. Assemble list of surfaces: pfTopo::addSurface(surf).
3. pfTopo::buildTopology().
4. pfTessellateAction::setBuildTopoWhileTess(FALSE).
5. pfdTessellateGeometry(shape, tessAction).
1. Create an pfTopo.
2. Assemble list of surfaces: pfTopo::addSurface().
3. Create pfBoundarys.
4. Add to list of boundaries: pfTopo::addBoundary().
5. Add edges to boundaries: pfBoundary::addEdge().
6. Set boundary orientation: pfEdge::setBoundaryDir().
%PFROOT\Src\pguide\libpf\C++
7. pfTessellateAction::setBuildTopoWhileTess(FALSE).
\topoTest.cxx
8. pfdTesselateGeometry(shape, tessAction).
(Microsoft Windows)
Build the topology “by hand.”
See the file
/usr/share/Performer/src/pgu
ide/libpf/C++/topoTest.C
(IRIX and Linux)
Step 7 does not appear in the code
because FALSE is the default.
007-1680-100
433
10: Creating and Maintaining Surface Topology
Reading and Writing Topology Information: Using Pseudo Loaders
You can add topological information to an existing set of connected, higher-order
surfaces in a file—for example, NURBS in an .iv or .csb file—and save the information
for future, crack-free surface rendering. As a result, you do not have to repeat the
topology build. The function pfdLoadFile() reads the topological information in a .pfb
file.
Before you save the scene graph data, you can also add tessellations that use the topology
to give crack-free images (see Chapter 11, “Rendering Higher-Order Primitives:
Tessellators”).
Table 10-2 shows three possible file conversions that you can apply to .iv or .csb files
that contain reps but no topology or tessellation; they are listed with example pfconv
command lines, which demonstrate how to use both the pfctol and pfttol pseudo
loaders.
Table 10-2
Adding Topology and Tessellations to .iv and .csb Files
Conversion
Example Command Line
Format change only.
pfconv sur.csb sur.pfb
Add topology information to scene
graph: save reps and topology
information but not tessellations.
pfconv sur.iv.topoTol.ttol surTopo.pfb
Add topology information and
tessellations to scene graph: save reps,
topology, and tessellations.
pfconv sur.iv.topoTol.ttol.tessTol.ctol surTopoTess.pfb
Add topology, tesselate, but do not
save reps.
pfconv sur.csb.topoTol.ttol.tessTol.ctol geodes.pfb.~.ctol
434
or
pfconv sur.csb.topoTol.ttol surTopo.pfb
or
pfconv sur.csb.topoTol.ttol.tessTol.ctol surTopoTess.pfb
007-1680-100
Summary of Scene Graph Topology: pfTopo
If you perform conversion, you may have files with or without tessellations. Depending
on the type of file you read, use one of the command lines in Table 10-3.
Table 10-3
Reading and Writing .pfb Files: with and without Tessellations
To read a .pfb file and perform
tessellation (without having to
build topology):
perfly surTopo.pfb.tessTol.ctol
To read a .pfb file that already
has tessellations
perfly surTopoTess.pfb
To read a .pfb file that already
has tessellations and force
retesselation (thus, removing any
existing geometry associated with
the higher order primitives)
perfly surTopoTess.pfb.+tessTol.ctol
To read a .pfb file that already
has tessellations and store it
without reps
perfly surTopoTess.pfb geodes.pfb.~.ctol
To delete the tessellation date, use the method clearTessellation().
Class Declaration for pfTopo
The class has the following main methods:
class pfTopo : public pfObject
{
public:
// Creating and destroying
pfTopo( );
virtual ~pfTopo();
// Accessor functions
void setDistanceTol( pfReal tol, pfLengthUnits u )
pfReal getDistanceTol( ) const;
pfLengthUnits getLengthUnits() const;
static pfTopo* getGlobalTopo(int n);
007-1680-100
435
10: Creating and Maintaining Surface Topology
static int getNumTopos();
pfParaSurface* getSurface( int i );
int getSurfaceCount( ) const;
pfBoundary* getBoundary( int i );
int getBoundaryCount( ) const;
int getSolidCount() const;
pfSolid* getSolid( int i )
//Adding topological elements
int addSurface( pfParaSurface *sur );
int addBoundary( pfBoundary
*bnd );
//Topology construction
void buildTopology();
int buildSolids();
};
Main Features of the Methods in pfTopo
buildSolids()
Collects connected surfaces in the pfTopo into pfSolids (see “Collecting
Connected Surfaces: pfSolid” on page 437).
buildTopology()
Builds consistent set of boundaries from the list of surfaces accumulated
by calls to addSurface(). Previously developed boundaries are deleted.
pfTopo(tol,u,sizeEstimate)
Construct a topological data structure.
tol specifies a tolerance for calculating when points are close enough
together to be considered the same. Default is 1 millimeter.
u specifies the system of units for tol. Default is meters.
getLengthUnits() and setLengthUnits()
Gets and sets the measurement units in object space for this pfTopo.
getNumTopos()
Returns the number of pfTopo structures in the global array of pfTopos.
getGlobalTopo()
Returns the specified pfTopo ID from the global array of pfTopos.
436
007-1680-100
Collecting Connected Surfaces: pfSolid
Note: In addition, it is possible to traverse the scene graph and build a consistent set of
boundaries for all the surfaces under a specified node by using the
pfdBuildTopologyTraverse(pfTopo *topo,pfNode *root) function.
The static member topology is an array of all topologies that have been created.
Collecting Connected Surfaces: pfSolid
To maintain consistent normals or propagate deformation information, organize
connected pfParaSurfaces in an pfSolid. With an pfSolid, you can collect connected
surface patches in one object for convenient access and manipulation.
Despite the name of the class, the set of surfaces need not form a closed surface, that is
the boundary of a volume. They can be a set of patches joined to form a surface, for
example, you might generate a hood of a car from two pfParaSurafaces that are mirror
images of each other.
To create solids, collect them in an pfTopo and then call pfTopo::buildSolid() (see
“Summary of Scene Graph Topology: pfTopo” on page 428).
Class Declaration for pfSolid
The class has the following main methods:
class pfSolid : public pfObject
{
public:
// Creating and destroying
pfSolid()
virtual ~pfSolid()
// Accessor functions
int addSurface( pfParaSurface *sur );
pfParaSurface* getSurface( int i);
int getSurfaceCount( ) const;
int getSolidId() const;
};
007-1680-100
437
10: Creating and Maintaining Surface Topology
Main Features of the Methods in pfSolid
Use the methods only after you have created an pfSolid with pfTopo::buildSolid().
Treat the method setSolidId() that appears in pfSolid.h as private: it is used by
pfTopo::buildSolid() when building the solid.
438
007-1680-100
Chapter 11
11. Rendering Higher-Order Primitives: Tessellators
To render a curve or surface, you must develop an approximation of it with pfGeoSets.
The tool that translates these reps into a mesh of contiguous triangles is called a tessellator.
Tessellation is interpretive; there is necessarily a difference between the original surface
and the tessellated mesh. You can control how closely you want the mesh to resemble the
surface.
•
Close resemblance, requiring many triangles, produces a realistic shape but incurs
slow graphic processing.
•
A gross approximation of the original surface results in fast processing.
Applications often create a series of tessellated representations of a shape, each one called
a level of detail (LOD). High resolution LODs are used when shapes are close to the
viewer and low resolution LODs are used when shapes are far from the viewer. Because
distance obscures detail, high resolution LODs are not necessary to represent distant
shapes.
This chapter describes how to control the tessellation of shapes in the following sections:
007-1680-100
•
“Features of Tessellators” on page 440
•
“Base Class pfTessellateAction” on page 443
•
“Tessellating Parametric Surfaces” on page 445
439
11: Rendering Higher-Order Primitives: Tessellators
pfObject
pfTessellateAction
pfTessParaSurfaceAction
Figure 11-1
Tessellators for
continuous
surfaces
Class Hierarchy for Tessellators
Features of Tessellators
Tessellators generate a sequence of straight-line segments to approximate an edge curve
of a surface, then cover the surface with triangular tiles. With each triangle vertex it
creates, a tessellator also stores the normal vector at the point from original surface. The
normal vectors are necessary for lighting and shading calculations.
Tessellations necessarily burden the entire graphics pipeline; they provide a first
definition of the rendering task by specifying a maximal set of vertices to be sent to the
graphics hardware.
440
007-1680-100
Features of Tessellators
Tessellators for Varying Levels of Detail
Ideally, you would quickly generate the simplest tessellation that adequately represents
surfaces of interest. What is adequate depends on your particular rendering task. You
may want to generate several tessellations with varying degrees of complexity and
accuracy for one pfRep and place them in level-of-detail nodes, as discussed in
“Level-of-Detail Management” in Chapter 5. The tessellators include accessor functions
to help you assess the load they create for the graphics hardware.
The control parameter for tessellations specifies the maximum deviation from the exact
surface. Figure 11-2 illustrates the effects of varying the deviation. The upper left image
is appropriate for accurate representation of the surface, the lower right image would be
appropriate if the object were in the distant background of a scene.
Figure 11-2
007-1680-100
Tessellations Varying With Changes in Control Parameter
441
11: Rendering Higher-Order Primitives: Tessellators
Details of Figure 11-2
The surface shown in Figure 11-2 was made with the repTest application using an
pfFrenetSweptSurface as follows (see “pfFrenetSweptSurface” on page 399):
pfReal profile( pfReal t ) { return 0.5*cos(t*6.0) + 1.25; };
pfSuperQuadCurve3d *cross =
new pfSuperQuadCurve3d( 0.75, new pfRVec3(0.0, 0.0, 0.0), 3.0
);
pfCircle3d *path = new pfCircle3d( 1.75, new pfRVec3(0.0, 0.0, 0.0) );
pfFrenetSweptSurface *fswept =
new pfFrenetSweptSurface( cross, path, profile );
fswept->setHandednessHint( true );
The number of triangles in Figure 11-2 decreases as the maximum-deviation parameter
chordalDevTol varies from .001 to .01 to .1 to .5 (see “Tessellating Parametric Surfaces” on
page 445). These numbers should be compared to the scale of the object, which has a
maximum diameter of 6.125 = 2(1.75 + 1.75 × .75), a minimum diameter of
.875 = 2(1.75 − 1.75 × .75), a maximum height of 2.625 = 2(1.75 × .75), and a minimum
height of 1.125 = 2(.75 × .75).
Tessellators Act on a Whole Graph or Single Node
You can apply a tessellator either to a scene graph or to just one node. The tessellators
produce a pfGeoSet from an pfRep and place that pfGeoSet in the pfGeode that holds the
pfRep.
Tessellators and Topology: Managing Cracks
A tessellation begins with a discrete set of vertices at surface edges. To prevent cracks
from appearing between adjacent surfaces, the same set of vertices should be used to
tessellate both surfaces.
To address the crack problem, you have several options, which are discussed in
“Building Topology: Computing and Using Connectivity Information” on page 431.
Table 10-1 on page 433 lists the different approaches to topology building, and the
methods to use for each.
442
007-1680-100
Base Class pfTessellateAction
Base Class pfTessellateAction
The pfTessellateAction class itself primarily stores statistics concerning the current
tessellation. The pfTessParaSurfaceAction class performs the tessellation of surfaces. You
can invoke the tessellation on a single surface by calling the following method:
pfTessParaSurfaceAction::tessellator(pfNode *n)
You can invoke the tessalltion on the entire scene graph by using the following function:
pfdTessellateGeometry(pfNode *root,pfTessParaSurfaceAction *tessAction)
Nodes which do not derive from pfParaSurface will be left untouched by the
pfdTessellateGeometry() function call.
Retessellating a Scene Graph
A tessellator will not tessellate a pfRep if the geoset count (pfGeode::getNumGSets()) is
not zero. If you want to retessellate a pfParaSurface, you must call
pfParaSurface::clearTessellation(). Note that you can also load a file and force
retessellation to occur by using the libpfctol pseudo loader and specifying the +
symbol in front of the ctol value you wish to use during tessellation, as shown in the
following:
perfly surfaces.pfb.+0.01.ctol
Class Declaration for pfTessellateAction
The class has the following main methods:
class pfTessellateAction : public pfObject
{
public:
// Creating and destroying
pfTessellateAction( void );
virtual ~pfTessellateAction( void );
// Accessor functions
void setExtSize( int s );
int getExtSize( );
int getTriangleCount() const;
int getTriStripCount() const;
007-1680-100
443
11: Rendering Higher-Order Primitives: Tessellators
int getTriFanCount() const;
void setReverseTrimLoop(const pfBool enable );
pfBool getReverseTrimLoop() const;
void setBuildTopoWhileTess(pfBool _buildTopoWhileTess);
pfBool getBuildTopoWhileTess() const;
void
setTopo(pfTopo * _topo);
pfTopo *getTopo( void )const;
};
Main Features of the Methods in pfTessellateAction
getTriangleCount()
Returns the number of all triangles generated by this instance of the
tessellator.
getTriStripCount() and getTriFanCount()
Return the number of tristrips or trifans in the tessellation.
setBuildTopoWhileTess() and getBuildTopoWhileTess()
Sets a flag whether surface connectivity is computed during the
tessellation traversal. Set the topology data structure to use with
setTopo().
If TRUE, before tessellating each surface, the connectivity of all
previously tessellated surfaces is used to avoid cracks when
tessellating. Notice that the final tessellations of the surfaces in the
scene graph may still have cracks because of unforeseen junctions
between surfaces.
If FALSE, no topology is constructed while tessellating. This leads to
two very different possible results:
•
444
If topology information for the surfaces to be tessellated was
developed before the tessellation, by calling
pfdBuildTopologyBuildTraverse() or pfTopo::buildTopology() or
by constructing topology by hand, the tessellator uses the
information and avoids cracks between surfaces. This option
provides the most crack-free tessellations possible.
007-1680-100
Tessellating Parametric Surfaces
•
If topology information was not developed before the tessellation
traversal, then surfaces are tessellated without regard to
connectivity and cracks appear between all adjacent surfaces. This
option provides the least crack-free tessellations possible.
setReverseTrimLoop() and getReverseTrimLoop()
Set and recover the orientation of trim loops. Recall that the side of the
surface to the left of the trim loop is rendered (see the section
“Parametric Surfaces” on page 375).
setTopo() and getTopo()
Set and get the pfTopo that holds the topology information used by the
tessellator (see “Summary of Scene Graph Topology: pfTopo” on
page 428).
Tessellating Parametric Surfaces
This section discusses the two classes OpenGL Performer provides for tessellating
parametric surfaces. The class pfTessParaSurfaceAction has methods for any parametric
surface.
pfTessParaSurfaceAction
The pfTessParaSurfaceAction class develops tessellations of any pfParaSurface. If a
surface has boundary curves, the tessellator starts there and specifies vertices at the
edges of the surface. The tessellator then covers the surface with pfGeoSets using the
boundary vertices to “pin” the edges of the tessellation. If necessary, the tessellator
creates edge vertices by constructing a discrete version of the boundary curve associated
with each of the surface’s pfEdges. An advantage of starting all tessellations at
boundaries is easy coordination of tessellations by several processors.
As part of the tessellation process, you can generate the UV coordinates for each vertex
created by the tessellator.
To control the accuracy of a tessellation, you specify a chordal deviation parameter which
constrains the distance of edges in the tessellation from the original surface.
007-1680-100
445
11: Rendering Higher-Order Primitives: Tessellators
Class Declaration for pfTessParaSurfaceAction
The class has the following main methods:
class pfTessParaSurfaceAction : public pfTessellateAction
{
public:
pfTessParaSurfaceAction();
pfTessParaSurfaceAction( pfReal chordalDevTol,
pfBool scaleTolByCurvature, int samples);
virtual ~pfTessParaSurfaceAction();
// Accessor functions
void
setChordalDevTol( const pfReal chordalDevTol );
pfReal getChordalDevTol( ) const;
void
setScaleTolByCurvature( const pfReal scaleTolByCurvature )
pfBool
getScaleTolByCurvature() const;
void setSampling( const int samples );
int getSampling( );
void setNonUniformSampling(const pfBool samples);
pfBool getNonUniformSampling() const;
void setGenUVCoordinates( const pfBool genUVCoordinates );
pfBool getGenUVCoordinates( ) const;
void setGenGeoArrays(pfBool enable);
pfBool getGenGeoArrays() const;
void tessellator(pfParaSurface *sur);
};
446
007-1680-100
Tessellating Parametric Surfaces
Main Features of the Methods in pfTessParaSurface
pfTessParaSurface()
Creates the class and provides a hint for the maximum deviation of the
tessellation from the original surface, indicates whether the tolerance
should be scaled by curvature, and provides a hint for how many
vertices to include in the tessellation.
setChordalDevTol() and getChordalDevTol()
Set and get the maximum distance from the original surface to the edges
produced by the tessellation.
setGenUVCoordinates() and getGenUVCoordinates()
Set and get a flag that indicates whether to generate UV coordinates for
the vertices produced in the tessellation. The coordinates for each vertex
are stored as the vertex’s texture coordinates.
setSampling() and getSampling()
Set and get the hint for the number of triangle vertices in the tessellation
along each boundary of the surface. If the surface has no trim curves
defining its “outer” edges, then the sampling is along the edges of the
UV rectangle that parameterizes the surface.
setScaleTolByCurvature() and getScaleTolByCurvature()
Set and get a flag to control whether the chordal deviation parameter
should be scaled by curvature. If nonzero, the tessellation of highly
curved areas improves.
setGenGeoArrays() and getGenGeoArrays()
Set and get a flag that determines whether or not the tessellator should
generate pfGeoArrays instead of pfGeoSets. By default, this is set to true.
tessellator()
Tessellates an individual surface.
Although pfTessParaSurface::tessellator() may be used to tessellate an individual
surface, it is more common to tessellate a set of surfaces using the
pfdTessellateGeometry() function.
007-1680-100
447
11: Rendering Higher-Order Primitives: Tessellators
Sample From repTest: Tessellating and Rendering a Sphere
The sample code in this section demonstrates how to create higher-level reps such as
pfParaSurfaces and pfCurve2ds and to save the output to a file. The lines of code perform
the following procedures:
•
Initializing OpenGL Performer
•
Creating and tessellating a pfSphereSurface
•
Saving the scene graph to a .pfb file
The code that follows can be found in the file
/usr/share/Performer/src/pguide/libpf/C++/repTest.C on IRIX and
Linux and %PFROOT%\Src\pguide\libpf\C++\repTest.cxx on Microsoft
Windows.
From main()
Initialize OpenGL Performer and
prepare for the creation of
higher-level reps.
pfInit();
pfdInitConverter(argv[1]?argv[1]:”pfb”);
shared = (SharedData *)pfCalloc(1,
sizeof(SharedData),
pfGetSharedArena());
shared->count = 10;
pfConfig();
pfGroup *root = new pfGroup();
Create Geometry
Create the geometry by calling
makeObjects(), which in turn calls
setupShape() to position the reps and
specify their geostate.
448
makeObjects(root);
007-1680-100
Tessellating Parametric Surfaces
Define setUpShape()
The function setupShape() creates a
new pfGeode, applies a pfGeoState,
and positions the pfRep using
pfRep::setOrigin().
static void
setupShape(pfGroup *p,
pfParaSurface *rep,pfReal x,
pfReal y,pfReal z)
{
// Get the current origin of the object
pfRVec3 org;
rep->getOrigin(org);
// Add the incoming offset to it
org[0] += x;
org[1] += y;
org[2] += z;
// Now reset the origin to include the
// incoming offset
rep->setOrigin(org);
// Set the appearance of this shape to
// be a random color
pfGeoState *gState
makeColor((float)rand()/((2<<15) - 1.0f),
(float)rand()/((2<<15) - 1.0f),
(float)rand()/((2<<15) - 1.0f));
// Set the geostate for the surface to
// be used during tessellation
rep->setGState(gState);
// Attach the rep to the scene graph
p->addChild(rep);
}
007-1680-100
449
11: Rendering Higher-Order Primitives: Tessellators
Define makeObjects()
The function makeObjects() sets up
the scene graph, defines the grid of
reps, and places the surfaces in the
scene graph by calling setupShape().
The code here shows the initial lines
of makeObjects() (omitting code that
controls the grid definition) and the
example of defining a trimmed and
untrimmed pfSphereSurface.
pfGroup *makeObjects(pfGroup *root)
{
....
pfSphereSurface *sphere = new
pfSphereSurface(3.0);
if(nVersions <= 0) {
pfCircle2d *trimCircle2d = new
pfCircle2d(1.0, pfRVec2(M_PI/2.0, M_PI));
sphere->addTrimCurve(0, trimCircle2d,
NULL);
}
setupShape(sphere, PF_XDIST*numObject++,
Y, PF_VIEWDIST);
....
}
450
007-1680-100
Chapter 12
12. Graphics State
The graphics state is a class of fields that defines everything about the shape and texture
of an object in an OpenGL Performer scene. Fields include such things as transparency,
shading, reflectance, and texture. The graphics state is set globally for all objects in the
scene graph. Individual objects, however, can override graphics state settings. The cost,
however, is efficiency. For performance reasons, therefore, it is important to set the fields
in the graphics state to satisfy the greatest number of objects in the scene.
This chapter describes in detail all of the fields in the graphics state.
Immediate Mode
The graphics libraries are immediate-mode state machines; if you set a mode, all
subsequent geometry is drawn in that mode. For the best performance, mode changes
need to be minimized and managed carefully. libpr manages a subset of graphics
library state and identifies bits of state as graphics state elements. Each state element is
identified with a PFSTATE token; for example., PFSTATE_TRANSPARENCY
corresponds to the transparency state element. State elements are loosely partitioned into
three categories: modes, values, and attributes.
Modes are the graphics state variables, such as transparency and texture enable, that
have simple values like ON and OFF. An example of a mode command is
pfTransparency(mode).
Values are not modal, rather they are real numbers which signify a threshold or quantity.
An example of a value is the reference alpha value specified with the pfAlphaFunc()
function.
Attributes are references to encapsulations (structures) of graphics state. They logically
group the more complicated elements of state, such as textures and lighting models.
Attributes are structures that are modified through a procedural interface and must be
007-1680-100
451
12: Graphics State
applied to have an effect. For example, pfApplyTex(tex) applies the texture map, tex, to
subsequently drawn geometry.
In libpr, there are three methods of setting a state:
•
Immediate mode
•
Display list mode
•
pfGeoState mode
Like the graphics libraries, libpr supports the notion of both immediate and display-list
modes. In immediate mode, graphics mode changes are sent directly to the Geometry
Pipeline; that is, they have an immediate effect. In display-list mode, graphics mode
changes are captured by the currently active pfDispList, which can be drawn later.
libpr display lists differ from graphics library objects because they capture only libpr
commands and are reusable. libpr display lists are useful for multiprocessing
applications in which one process builds up the list of visible geometry and another
process draws it. “Display Lists” on page 479 describes libpr display lists.
A pfGeoState is a structure that encapsulates all the graphics modes and attributes that
libpr manages. You can individually set the state elements of a pfGeoState to define a
graphics context. The act of applying a pfGeoState with pfApplyGState() configures the
state of the Geometry Pipeline according to the modes, values, and attributes set in the
pfGeoState. For example, the following code fragment shows equivalent ways (except
for some inheritance properties of pfGeoStates described later) of setting up some
lighting parameters suitable for a glass surface:
/* Immediate mode state specification */
pfMaterial *shinyMtl;
pfTransparency(PFTR_ON);
pfApplyMtl(shinyMtl);
pfEnable(PFEN_LIGHTING);
/* is equivalent to: */
/* GeoState state specification */
pfGeoState *gstate;
pfGStateMode(gstate, PFSTATE_TRANSPARENCY, PFTR_ON);
pfGStateAttr(gstate, PFSTATE_FRONTMTL, shinyMtl);
pfGStateMode(gstate, PFSTATE_ENLIGHTING, PF_ON);
pfApplyGState(gstate);
452
007-1680-100
Immediate Mode
In addition, pfGeoStates have unique state inheritance capabilities that make them very
convenient and efficient; they provide independence from ordered drawing. pfGeoStates
are described in the section “pfGeoState” on page 482 of this chapter.
The libpr routines have been designed to produce an efficient structure for managing
graphics state. You can also set a graphics state directly through the GL. However, libpr
will have no record of these settings and will not be able to optimize them and may make
incorrect assumptions about current graphics state if the resulting state does not match
the libpr record when libpr routines are called. Therefore, it is best to use the libpr
routines whenever possible to change a graphics state and to restore the libpr state if
you go directly through the GL.
The following sections will describe the rendering geometry and state elements in detail.
There are three types of state elements: modes, values, and attributes. Modes are simple
settings that take a set of integer values that include values for enabling and disabling the
mode. Modes may also have associated values that allow a setting from a defined range.
Attributes are complex state structures that encapsulate a related collection of modes and
values. Attribute structures will not include in their definition an enable or disable as the
enabling or disabling of a mode is orthogonal to the particular related attribute in use.
Rendering Modes
The libpr library manages a subset of the rendering modes found in OpenGL. In
addition, libpr abstracts certain concepts like transparency, providing a higher-level
interface that hides the underlying implementation mechanism.
The libpr library provides tokens that identify the modes that it manages. These tokens
are used by pfGeoStates and other state-related functions like pfOverride(). The
following table enumerates the PFSTATE_* tokens of supported modes, each with a brief
description and default value.
Table 12-1 lists and describes the mode tokens.
Table 12-1
007-1680-100
pfGeoState Mode Tokens
Token Name
Description
Default Value
PFSTATE_TRANSPARENCY
Transparency modes
PFTR_OFF
PFSTATE_ALPHAFUNC
Alpha function
PFAF_ALWAYS
453
12: Graphics State
Table 12-1
pfGeoState Mode Tokens (continued)
Token Name
Description
Default Value
PFSTATE_ANTIALIAS
Antialiasing mode
PFAA_OFF
PFSTATE_CULLFACE
Face culling mode
PFCF_OFF
PFSTATE_DECAL
Decaling mode for coplanar geometry
PFDECAL_OFF
PFSTATE_SHADEMODEL
Shading model
PFSM_GOURAUD
PFSTATE_ENLIGHTING
Lighting enable flag
PF_OFF
PFSTATE_ENTEXTURE
Texturing enable flag
PF_OFF
PFSTATE_ENFOG
Fogging enable flag
PF_OFF
PFSTATE_ENWIREFRAME
pfGeoSet wireframe mode enable flag
PF_OFF
PFSTATE_ENCOLORTABLE
pfGeoSet colortable enable flag
PF_OFF
PFSTATE_ENHIGHLIGHTING
pfGeoSet highlighting enable flag
PF_OFF
PFSTATE_ENLPOINTSTATE
pfGeoSet light point state enable flag
PF_OFF
PFSTATE_ENTEXGEN
Texture coordinate generation enable flag PF_OFF
PFSTATE_ENFRAGPROG
Fragment program enable flag
PF_OFF
PFSTATE_ENVTXPROG
Vertex program enable flag
PF_OFF
PFSTATE_ENSHADPROG
Shader program enable flag
PF_OFF
The mode control functions described in the following sections should be used in place
of their graphics library counterparts so that OpenGL Performer can correctly track the
graphics state. Use pfGStateMode() with the appropriate PFSTATE token to set the
mode of a pfGeoState.
454
007-1680-100
Immediate Mode
Transparency
You can control transparency using pfTransparency(). Possible transparency modes are
listed in the following table.
Table 12-2
pfTransparency Tokens
Transparency mode
Description
PFTR_OFF
Transparency disabled.
PFTR_ON
PFTR_FAST
Use the fastest, but not necessarily the best, transparency provided
by the hardware.
PFTR_HIGH_QUALITY
Use the best, but not necessarily the fastest, transparency provided
by the hardware.
PFTR_MS_ALPHA
Use screen-door transparency when multisampling. Fast but limited
number of transparency levels.
PFTR_BLEND_ALPHA
Use alpha-based blend with background color. Slower but high
number of transparency levels.
In addition, the flag PFTR_NO_OCCLUDE may be logically ORed into the transparency
mode in which case geometry will not write depth values into the frame buffer. This will
prevent it from occluding subsequently rendered geometry. Enabling this flag improves
the appearance of unordered, blended transparent surfaces.
There are two basic transparency mechanisms: screen-door transparency, which requires
hardware multisampling, and blending. Blending offers very high-quality transparency
but for proper results requires that transparent surfaces be rendered in back-to-front
order after all opaque geometry has been drawn. When using transparent texture maps
to “etch” geometry or if the surface has constant transparency, screen-door transparency
is usually good enough. Blended transparency is usually required to avoid “banding” on
surfaces with low transparency gradients like clouds and smoke.
Shading Model
You can select either flat shading or Gouraud (smooth) shading. pfShadeModel() takes
one of two tokens: PFSM_FLAT or PFSM_GOURAUD. One some graphics hardware flat
shading can offer a significant performance advantage.
007-1680-100
455
12: Graphics State
Alpha Function
The pfAlphaFunc() function is an extension of the glAlphaFunc() function; it allows
OpenGL Performer to keep track of the hardware mode. The alpha function is a pixel test
that compares the incoming alpha to a reference value and uses the result to determine
whether or not the pixel is rendered. The reference value must be specified in the range
[0, 1]. For example, a pixel whose alpha value is 0 is not rendered if the alpha function is
PFAF_GREATER and the alpha reference value is also 0. Note that rejecting pixels-based
alpha can be faster than using transparency alone. A common technique for improving
the performance of filling polygons is to set an alpha function that will reject pixels of low
(possibly nonzero) contribution. The alpha function is typically used for see-through
textures like trees.
Decals
On Z-buffer-based graphics hardware, coplanar geometry can cause unwanted artifacts
due to the finite numerical precision of the hardware which cannot accurately resolve
which surface has visual priority. This can result in flimmering, a visual “tearing” or
“twinkling” of the surfaces. pfDecal() is used to accurately draw coplanar geometry on
SGI graphics platforms and it supports two basic implementation methods : stencil
decaling and displace decaling.
The stencil decaling method uses a hardware resource known as a stencil buffer and
requires that a single stencil plane (see the man page for glStencilOp()) be available for
OpenGL Performer. This method offers the highest image quality but requires that
geometry be coplanar and rendered in a specific order which reduces opportunities for
the performance advantage of sorting by graphics mode.
A potentially faster method is the displace decaling method. In this case, each layer is
displaced towards the eye so it hovers slightly above the preceding layer. Displaced
decals need not be coplanar, and can be drawn in any orde, but the displacement may
cause geometry to incorrectly poke through other geometry.
The specificaton of a decal plane can improve the displace decaling method. The object
geometry will be projected onto the specified plane and if the same plane is specified for
base and layer geometry, the base and layer polygons will be generated with identical
depth values. If the objects are drawn in priority order, no further operation is necessary.
Otherwise, displace can be applied to the planed geometry for a superior result. Decal
planes can be specified on pfGeoSets with pfGSetDecalPlane(), on a pfGeoState with the
PFSTATE_DECALPLANE attribute to pfGStateAttr(), or globally with
pfApplyDecalPlane().
456
007-1680-100
Immediate Mode
Decals consist of base geometry and layer geometry. The base defines the depth values of the
decal while layer geometry is simply inlaid on top of the base. Multiple layers are
supported but limited to eight when using displaced decals. Realize that these layers
imply superposition; there is no limit to the number of polygons in a layer, only to the
number of distinct layers.
The decal mode indicates whether the subsequent geometry is base or layer and the decal
method to use. For example, a mode of PFDECAL_BASE_STENCIL means that
subsequent geometry is to be considered as base geometry and drawn using the stencil
method. All combinations of base/layer and displace/stencil modes are supported but
you should make sure to use the same method for a given base-layer pair.
Example 12-1 illustrates the use of pfDecal().
Example 12-1
Using pfDecal() to a Draw Road with Stripes
pfDecal(PFDECAL_BASE_STENCIL);
/* ... draw underlying geometry (roadway) here ...*/
pfDecal(PFDECAL_LAYER_STENCIL);
/* ... draw coplanar layer geometry (stripes) here ... */
pfDecal(PFDECAL_OFF);
Note: libpf applications can use the pfLayer node to include decals within a scene
graph.
Frontface / Backface
The pfCullFace() function controls which side of a polygon (if any) is discarded in the
Geometry Pipeline. Polygons are either front-facing or back-facing. A front-facing
polygon is described by a counterclockwise order of vertices in screen coordinates, and
a back-facing one has a clockwise order. pfCullFace() has four possible arguments:
007-1680-100
PFCF_OFF
Disable face-orientation culling.
PFCF_BACK
Cull back-facing polygons.
PFCF_FRONT
Cull front-facing polygons.
457
12: Graphics State
PFCF_BOTH
Cull both front- and back-facing polygons.
In particular, back-face culling is highly recommended since it offers a significant
performance advantage for databases where polygons are never be seen from both sides
(databases of “solid” objects or with constrained eyepoints).
Antialiasing
The pfAntialias() function is used to turn the antialiasing mode of the hardware on or
off. Currently, antialiasing is implemented differently by each different graphics system.
Antialiasing can produce artifacts as a result of the way OpenGL Performer and the
active hardware platform implement the feature. See the man page for pfAntialias() for
implementation details.
Rendering Values
Some modes may also have associated values. These values are set through
pfGStateVal(). Table 12-3 lists and describes the value tokens.
Table 12-3
pfGeoState Value Tokens
Token Name
Description
Range
PFSTATE_ALPHAREF
Set the alpha function reference value. 0.0 - 1.0
Default Value
0.0
Enable / Disable
The pfEnable() and pfDisable() functions control certain rendering modes. Certain
modes do not have effect when enabled but require that other attribute(s) be applied.
458
007-1680-100
Immediate Mode
Table 12-4 lists and describes the tokens and also lists the attributes required for the mode
to become truly active.
Table 12-4
Enable and Disable Tokens
Token
Action
Attribute(s) Required
PFEN_LIGHTING
Enable or disable lighting.
pfMaterial
pfLight
pfLightModel
PFEN_TEXTURE
Enable or disable texture.
pfTexEnv
pfTexture
PFEN_FOG
Enable or disable fog.
pfFog
PFEN_WIREFRAME
Enable or disable pfGeoSet wireframe rendering. none
PFEN_COLORTABLE
Enable or disable pfGeoSet colortable mode.
pfColortable
PFEN_HIGHLIGHTING
Enable or disable pfGeoSet highlighting.
pfHighlight
PFEN_TEXGEN
Enable or disable automatic texture coordinate
generation.
pfTexGen
PFEN_LPOINTSTATE
Enable or disable pfGeoSet light points.
pfLPointState
PFEN_FRAGPROG
Enable or disable fragment program.
pfFragmentProgram
PFEN_VTXPROG
Enable or disable vertex program.
pfVertexProgram
PFEN_SHADPROG
Enable or disable shader programs.
pfShaderProgram
By default all modes are disabled.
Rendering Attributes
Rendering attributes are state structures that are manipulated through a procedural
interface. Examples include pfTexture, pfMaterial, and pfFog. libpr provides tokens
that enumerate the graphics attributes it manages. These tokens are used by pfGeoStates
007-1680-100
459
12: Graphics State
and other state-related functions like pfOverride(). Table 12-5 lists and describes the
tokens.
Table 12-5
Rendering Attribute Tokens
Attribute Token
Description
Apply Routine
PFSTATE_LIGHTMODEL
Lighting model
pfApplyLModel()
PFSTATE_LIGHTS
Light source definitions
pfLightOn()
PFSTATE_FRONTMTL
Front-face material
pfApplyMtl()
PFSTATE_BACKMTL
Back-face material
pfApplyMtl()
PFSTATE_TEXTURE
Texture
pfApplyTex()
PFSTATE_TEXENV
Texture environment
pfApplyTEnv()
PFSTATE_FOG
Fog model
pfApplyFog()
PFSTATE_COLORTABLE
Color table for pfGeoSets
pfApplyCtab()
PFSTATE_HIGHLIGHT
pfGeoSet highlighting style
pfApplyHlight()
PFSTATE_LPOINTSTATE pfGeoSet light point definition
pfApplyLPState()
PFSTATE_TEXGEN
Texture coordinate generation
pfApplyTGen()
PFSTATE_FRAGPROG
Fragment program definition
pfGProgramApply()
PFSTATE_VTXPROG
Vertex program definition
pfGProgramApply()
PFSTATE_GPROGPARMS
GPU program parameters definition
pfGPParamsApply()
PFSTATE_SHADPROG
Shader program definition
pfShaderProgramApply()
Rendering attributes control which attributes are applied to geometric primitives when
they are processed by the hardware. All OpenGL Performer attributes consist of a control
structure, definition routines, and an apply function, pfApply* (except for lights which
are “turned on”).
Each attribute has an associated pfNew*() routine that allocates storage for the control
structure. When sharing attributes across processors in a multiprocessor application, you
should pass the pfNew*() routine a shared memory arena from which to allocate the
structure. If you pass NULL as the arena, the attribute is allocated from the heap and is
not sharable in a nonshared address space (fork()) multiprocessing application.
460
007-1680-100
Immediate Mode
All attributes can be applied directly, referenced by a pfGeoState or captured by a display
list. When changing an attribute, that change is not visible until the attribute is reapplied.
Detailed coverage of attribute implementation is available in the man pages.
Texture
OpenGL Performer supports texturing through pfTextures and pfTexEnvs, which
provide encapsulated suppport for graphics library textures (see glTexImage2D() ) and
texture environments (see glTexEnv()). A pfTexture defines a texture image, format, and
filtering. A pfTexEnv specifies how the texture should interact with the colors of the
geometry where it is applied. You need both to display textured data, but you do not
need to specify them both at the same time. For example, you could have pfGeoStates
each of which had a different texture specified as an attribute and still use an overall
texture environment specified with pfApplyTEnv().
A pfTexture is created by calling pfNewTex(). If the desired texture image exists as a disk
file in IRIS RGB image format (the file often has a “.rgb” suffix ) or in the OpenGL
Performer fast-loading image format (a “.pfi” suffix), you can call pfLoadTexFile() to
load the image into CPU memory and completely configure the pfTexture, as shown in
the following:
pfTexture *tex = pfLoadTexFile(“brick.rgba”);
Otherwise, pfTexImage() lets you directly provide a GL-ready image array in the same
external format as specified on the pfTexture and as expected by glTexImage2D(), as
shown in the following:
void pfTexImage(pfTexture* tex, uint* image,
int comp, int sx, int sy, int sz);
OpenGL expects packed texture data with each row beginning on a long word boundary.
However, OpenGL expects the individual components of a texel to be packed in opposite
order. For example, OpenGL expects the texels to be packed as RGBA. If you provide
your own image array in a multiprocessing environment, it should be allocated from
shared memory (along with your pfTexture) to allow different processes to access it. A
basic example demonstrating loading a image file and placing the resulting pfTexture on
scene graph geometry is at
/usr/share/Performer/src/pguide/libpf/C/texture.c on IRIX and Linux
and %PFROOT%\Src\pguide\libpf\C\texture.c on Microsoft Windows.
007-1680-100
461
12: Graphics State
Note: The size of your texture must be an integral power of two on each side. OpenGL
refuses to accept badly sized textures. You can rescale your texture images with the
izoom or imgworks programs (shipped with IRIX 5.3 in the eoe2.sw.imagetools and
imgtools.sw.tools subsystems; and with IRIX 6.2 in the eoe.sw.imagetools and
imgworks.sw.tools subsystems).
Your texture source does not have to be a static image. pfTexLoadMode() can be used to
set one of the sources listed in Table 12-6 with PFTEX_LOAD_SOURCE. Note that sources
other than CPU memory may not be supported on all graphics platforms, or may have
some special restrictions. There are several sample programs that demonstrate paging
sequences of texture from different texture sources. For paging from host memory there
are the following:
•
•
On IRIX and Linux:
–
/usr/share/Performer/src/pguide/libpr/C/texlist.c
–
/usr/share/Performer/src/pguide/libpr/C/mipmap.c
–
/usr/share/Performer/src/pguide/libpfutil/movietex.c
On Microsoft Windows:
–
%PFROOT%\Src\pguide\libpr\C\texlist.c
–
%PFROOT%\Src\pguide\libpr\C\mipmap.c
–
%PFROOT%\Src\pguide\libpfutil\movietex.c
These examples demonstrate the use of different texture sources for paging textures, and
libpufitl utilties for managing texture resources. One thing these examples do is use
pfTexLoadImage() to update the pointer to the image data to avoid the expensive
reformating the texture. This requires that the provided image data be the same size as
the original image data, as well as same number of components and same formats.
Table 12-6
Texture Image Sources
PFTEX_SOURCE_ Token Source of the Texture Image
IMAGE
CPU memory location specified by pfTexLoadImage() or
pfTexImage()
462
007-1680-100
Immediate Mode
Table 12-6
Texture Image Sources (continued)
PFTEX_SOURCE_ Token Source of the Texture Image
FRAMEBUFFER
Framebuffer location offset from window origin as specified by
pfTexLoadOrigin()
VIDEO
Default video source on the system
Video Texturing
The source of texture image data can be live video input. OpenGL Performer supports
the video input mechanisms of Sirius Video on RealityEngine and InfiniteReality, DIVO
on InfiniteReality, and Silicon Graphics O2 and Octane video input. OpenGL Performer
includes a sample program that features video texturing:
/usr/share/Performer/src/pguide/libpf/movietex.c (IRIX and Linux)
%PFROOT%\Src\pguide\libpf\movietex.c (Microsoft Windows)
This example demonstrates that all video initialization, including the creation of video
library resources, is done in the draw process, as required by the video library. Part of that
initialization includes setting the proper number of components on the pfTexture and
choosing a texture filter and potentially internal format. Those basic operations are
discussed further in this section.
OpenGL Performer will automatically download the frame of video when the texture
object is applied through the referencing pfGeoState. Alternatively, you may want to
schedule this download to happen at the beginning or end of the rendering frame; you
can force it with pfLoadTex().
Textures must be created with sizes that are powers of 2; the input video frame is usually
not in powers of 2. The [0,1] texture coordinate range can be scaled into the valid part of
the pfTexture with a texture matrix. This matrix can be applied directly to the global state
with pfApplyTMat() to affect all pfGeoSets, or can be set on a pfGeoState with the
PFSTATE_TEXMAT attribute to pfGStateAttr().
Texture Management
Texture storage is limited only by virtual memory, but for real-time applications you
must consider the amount of texture storage the graphics hardware supports. Textures
that do not fit in the graphics subsystem are paged as needed when pfApplyTex() is
called. The libpr library provides routines for managing hardware texture memory so
007-1680-100
463
12: Graphics State
that a real-time application does not have to get a surprise texture load. pfIsTexLoaded(),
called from the drawing process, tells you if the pfTexture is currently properly loaded in
texture memory. The initially required textures of an application, or all of the textures if
they fit, can be preloaded into texture memory as part of application initialization.
pfuMakeSceneTexList() makes a list of all textures referenced by a scene graph and
pfuDownloadTexList() loads a list of textures into hardware texture memory (and must
be called in the draw process). The Perfly sample program does this as part of its
initialization and displays the textures as it preloads them.
There are additional routines to assist with the progressive loading and unloading of
pfTextures. pfIdleTex() can be used to free the hardware texture memory owned by a
pfTexture. GL host and hardware texture memory resources can be freed with
pfDeleteGLHandle() from the drawing process. OpenGL Performer will automatically
re-allocate those resources if the pfTexture is used again. For an example of management
of texture resources, see the example program
/usr/share/Performer/src/pguide/libpfutil/texmem.c for IRIX and Linux
or %PFROOT%\Src\pguide\libpfutil\texmem.c for Microsoft Windows. The
program uses the pfuTextureManager from libpfutil for basic texture paging
support.
The pfLoadTex() function, called from the drawing process, can be used to explicitly load
a texture into graphics hardware texture memory (which includes doing any necessary
formatting of the texture image). By default, pfLoadTex() loads the entire texture image,
including any required minification or magnification levels, into texture memory.
pfSubloadTex() and pfSubloadTexLevel() can also be used in the drawing process to do
an immediate load of texture memory managed by the given pfTexture and these
routines allow you to specify all loading parameters (source, origin, size, etc.). This is
useful for loading different images for the same pfTexture in different graphics pipelines.
pfSubloadTex() allows you to load a subsection of the texture tile by tile.
A special pfTexFormat() formatting mode, PFTEX_SUBLOAD_FORMAT, allows part or all
of the image in texture memory owned by the pfTexture to be replaced using
pfApplyTex(), pfLoadTex(), or pfSubloadTex(), without having to go through the
expensive reformatting phase. This allows you to quickly update the image of a
pfTexture in texture memory. The PFTEX_SUBLOAD_FORMAT used with an appropriate
pfTexLoadSize() and pfTexLoadOrigin() allows you to control what part of the texture
will be loaded by subsequent calls to pfLoadTex() or pfApplyTex(). There are also
different loading modes that cause pfApplyTex() to automatically reload or subload a
texture from a specified source. If you want the image of a pfTexture to be updated upon
every call to pfApplyTex(), you can set the loading mode of the pfTexture with
pfTexLoadMode() to be PFTEX_BASE_AUTO_REPLACE. pfTexLoadImage() allows you
464
007-1680-100
Immediate Mode
to continuously update the memory location of an IMAGE source texture without
triggering any reformatting of the texture.
Hint: There are texture formatting modes that can improve texture performance and
these are the modes that are used by default by OpenGL Performer. Of most importance
is the 16-bit texel internal formats. These formats cause the resulting texels to have 16 bits
of resolution instead of the standard 32. These formats can have dramatically faster
texture-fill performance and cause the texture to take up half the hardware texture
memory. Therefore, they are strongly recommended and are used by default. There are
different formats for each possible number of components to give a choice of how the
compression is to be done. These formats are described in the pfTexFormat(3pf) man
page.
There may also be formatting modes for internal or external image formats for which
OpenGL Performer does not have a token. However, the GL value can be specified.
Specifying GL values will make your application GL-specific and may also cause future
porting problems; so, it should only be done if absolutely necessary.
The pfTexture class also allows you to define a set of textures that are mutually exclusive;
they should always be applied to the same set of geometry; and, thus, they can share the
same location in hardware texture memory. With pfTexList(tex, list) you can specify a list
of textures to be in a texture set managed by the base texture, tex. The base texture is what
gets applied with pfApplyTex(), or assigned to geometry through pfGeoStates. With
pfTexFrame(), you can select a given texture from the list (–1 selects the base texture and
is the default). This allows you to define a texture movie where each image is the frame
of the movie. You can have an image on the base texture to display when the movie is not
playing. There are additional loading modes for pfTexLoadMode() described in
007-1680-100
465
12: Graphics State
Table 12-7 to control how the textures in the texture list share memory with the base
texture.
Table 12-7
Texture Load Modes
PFTEX_LOAD_
modeToken
Load Mode Values
Description
SOURCE
SOURCE_IMAGE,
SOURCE_FRAMEBUFFER
PFTEX_SOURCE_VIDEO
PFTEX_SOURCE_DMBUF
PFTEX_SOURCE_DMVIDEO
Source of image data is host memory image,
framebuffer, default video source, digital
media buffer, or a video library digital media
buffer.
BASE
BASE_APPLY
BASE_AUTO_SUBLOAD
Loading of image for pfTexture is done as
required for pfApply, or automatically
subloaded upon every pfApply().
LIST
LIST_APPLY
LIST_AUTO_IDLE
LIST_AUTO_SUBLOAD
Loading of list texture image is done as
separate apply, causes freeing of previous list
texture in hardware texture memory, or is
subloaded into memory managed by the base
texture.
VIDEO_
INTERLACED
OFF, INTERLACED_ODD,
INTERLACED_EVEN,
Video input is interlaced or not and if so,
which field (even or odd) is spatially higher.
Texture list textures can share the exact graphics texture memory as the base texture but
this has the restriction that the textures must all be the exact same size and format as the
base texture. Texture list textures can also indicate that they are mutually exclusive,
which will cause the texture memory of previous textures to be freed before applying the
new texture. This method has no restrictions on the texture list, but is less efficient than
the previous method. Finally, texture list textures can be treated as completely
independent textures that should all be kept resident in memory for rapid access upon
their application.
The pfTexFilter() function sets a desired filter to a specified filtering method on a
pfTexture. The minification and magnification texture filters are described with bitmask
tokens. If filters are partially specified, OpenGL Performer fills the rest with
machine-dependent fast defaults. The PFTEX_FAST token can be included in the
bitmask to allow OpenGL Performer to make machine dependent substitutions where
there are large performance differences.
466
007-1680-100
Immediate Mode
There are a variety of texture filter functions that can improve the look of textures when
they are minified and magnified. By default, textures use MIPmapping when minified
(though this costs an extra 1/3 in storage space to store the minification levels). Each level
of minification or magnification of a texture is twice the size of the previous level.
Minification levels are indicated with positive numbers and magnification levels are
indicated with nonpositive numbers. The default magnification filter for textures is
bilinear interpolation. The use of detail textures and sharpening filters can improve the
look of magnified textures. Detailing actually uses an extra detail texture that you
provide that is based on a specified level of magnification from the corresponding base
texture. The detail texture can be specified with the pfTexDetail() function. By default,
MIPmap levels are generated for the texture automatically. OpenGL operation allows for
the specification of custom MIPmap levels. Both MIPmap levels and detail levels can be
specified with pfTexLevel(). The level number should be a positive number for a
minification level and a nonpositive number for a magnification (detail) level. If you are
providing your own minification levels, you must provide all log2(MAX(texSizeX,
texSizeY)) minification levels. There is only one detail texture for a pfTexture.
Standard MIPmap filtering can induce blurring on a texture if the texture is applied to a
polygon which is angled away from the viewer. To reduce this blurring, an anisotropic
filter can be used to improve visual quality. pfTexAnisotropy() sets the degree of
anisotropy to be used by the specified pfTexture. The default degree of anisotropy is 1,
which is the same as the standard isotropic filter. A value of 2 will apply a 2:1 anisotropic
filter. The maximum degree of anisotropy can be queried with pfQuerySys().
Anisotropic filtering is supported on OpenGL implementations that support the
GL_EXT_texture_filter_anisotropic extension. pfQueryFeature() can be used
to determine if anisotropic filtering is supported on the current platform. If the
environment variable PF_MAX_ANISOTROPY is set, then an anisotropic filter of the value
specified by PF_MAX_ANISOTROPY will be applied to pfTextures that do not set the
degree of anisotropy.
The magnification filters use spline functions to control their rate of application as a
function of magnification and specified level of magnification for detail textures. These
splines can be specified with pfTexSpline(). The specification of the spline is a set of
control points that are pairs of nondecreasing magnification levels (specified with
nonpositive numbers) and corresponding scaling factors. Magnification filters can be
applied to all components of a texture, only the RGB components of a texture, or to just
the alpha components. OpenGL does not allow different magnification filters (between
detail and sharpen) for RGB and alpha channels.
007-1680-100
467
12: Graphics State
Note: The specification of detail textures may have GL dependencies and magnifications
filters may not be available on all hardware configurations. The pfTexture man page
describes these details.
Texture Formats
The format in which an image is stored in texture memory is defined with pfTexFormat():
void pfTexFormat(pfTexture *tex, int format, int type)
The format variable specifies which format to set. Valid formats and their basic types
include the following:
•
PFTEX_INTERNAL_FORMAT— Specifies how many bits per component are to be
used in internal hardware texture memory storage. The default is 16 bits per full
texel and is based on the number of components and external format. See the
pfTexture man page for the list of supported formats. Floating point formats are
supported only on selected platforms (for example, Onyx4 and Prism systems).
•
PFTEX_IMAGE_FORMAT—Describes the type of image data and must match the
number of components, such as PFTEX_LUMINANCE, PFTEX_LUMINANCE_ALPHA,
PFTEX_RGB, and PFTEX_RGBA. The default is the token in this list that matches the
number of components. Other OpenGL selections can be specified with the GL
token.
•
PFTEX_EXTERNAL_FORMAT—Specifies the format of the data in the pfTexImage
array. The default is packed 8 bits per component. There are special, fast-loading
hardware-ready formats, such as PFTEX_UNSIGNED_SHORT_5_5_5_1.
•
PFTEX_SUBLOAD_FORMAT—Specifies if the texture will be a subloadable paging
texture. The default is FALSE.
•
PFTEX_CUBE_MAP—Specifies that the texture is a cube map texture that contains
six images. The default is FALSE. Cube maps are supported only on selected
platforms (for example, Onyx4 or Prism systems).
In the case of cube maps, where there are six images, the images are specified using the
function pfTexMultiImage(). The parameter imageIndex with value 0–5 specifies the
cube face in the following order:
min_x, max_x, min_y, max_y, min_z, max_z
468
007-1680-100
Immediate Mode
For the order of faces and how the vector from the cube center to each texel is computed,
see the sample program in following file:
/usr/share/Performer/src/pguide/libpf/C++/cubeMap.C
(IRIX and Linux)
%PFROOT%\Src\pguide\libpf\C++\cubeMap.C
(Microsoft Windows)
All six images have to be specified, they must have the same size, and the value ns has
to be equal to nt.
The following are other variants of functions that are used by the cube maps:
•
pfTexMultiName()
•
pfGetTexMultiName()
•
pfLoadMultiTexFile()
•
pfSaveMultiTexFile()
•
pfSubloadMultiTex()
•
pfSubloadTexMultiLevel()
Other than having six images, the cube maps are used as any other pfTexture. The
exception is that subloads from other sources than user specified memory are not
supported.
In general, you will just need to specify the number of components in pfTexImage(). You
may want to specify a fast-loading hardware-ready external format, such as
PFTEX_UNSIGNED_SHORT_5_5_5_1, in which case OpenGL Performer automatically
chooses a matching internal format. See the pfTexFormat(3pf) man page for more
informaton on texture configuration details.
Controlling Texture LOD with pfTexLOD
You can control the levels of detail (LODs) of a texture that are accessed and used with
pfTexLOD to force higher or lower MIPmap levels to be used when minifying. You can
use this to give the graphics hardware a hint about what levels can be accessed (Impact
hardware takes great advantage of such a hint) and you can use this to have multiple
textures sharing a single MIPmap pyramid in texture memory. For example, a distant
object and a close one may use different LODs of the same pfTexture texture. The
pfGeoStates of those pfGeoSets would have different pfTexLOD objects that referenced
007-1680-100
469
12: Graphics State
the proper texture LODs. pfTexLevel() would be used to specify and update the proper
image for each LOD in the pfTexture. You can use LODs to specify to yourself and the GL
which LODs of texture should be loaded from disk into main memory. For example, if
the viewer is in one LOD, most of the texture in that LOD can often be viewed and,
consequently, should be paged into texture memory. You can set LOD parameters on a
pfTexture directly or use pfTexLOD.
To use a pfTexLOD object, you do the following:
1.
Set the ranges of the LOD using pfTLODRange() and their corresponding
minimum and maximum resolution MIPmap. Because the minimum and maximum
limits can be floating-point values, new levels can be smoothly blended in when
they become available to avoid popping from one LOD to another.
2. Optionally, set the bias levels using pfTLODBias() to force blurring of a texture to
simulate motion blur and depth of field, to force a texture to be sharper, or to
compensate for asymmetric minification of a MIPmapped texture.
Note: Any LOD settings on pfTexture take priority over current pfTexLOD settings.
3. Enable LOD control over texture by using the following methods:
pfEnable(PFEN_TEXLOD);
pfGeoState::pfGStateMode(myTxLOD, PFSTATE_ENTEXLOD, ON);
where myTxLOD is an instance of pfTexLOD, and ON is a nonzero integer.
4. Apply the LOD settings to the texture using pfApplyTLOD().
See the following sample program for an example of using a pfTexLOD:
/usr/share/Performer/src/pguide/libpr/C/texlod.c (IRIX and Linux)
%PFROOT%\Src\pguide\libpr\C\texlod.con (Microsoft Windows)
Setting the Texture Environment with pfTexEnv
The environment specifies how the colors of the geometry, potentially lit, and the texture
image interact. This is described with a pfTexEnv object. The mode of interaction is set
with pfTEnvMode() and valid modes include the following:
•
470
PFTE_MODULATE—The gray scale of the geometry is mixed with the color of the
texture (the default).
007-1680-100
Immediate Mode
This option multiplies the shaded color of the geometry by the texture color. If the
texture has an alpha component, the alpha value modulates the geometry’s
transparency; for example, if a black and white texture, such as text, is applied to a
green polygon, the polygon remains green and the writing appears as dark green
lettering.
•
PFTE_DECAL—The texture alpha component acts as a selector between 1.0 for the
texture color and 0.0 for the base color to decal an image onto geometry.
•
PFTE_BLEND—The alpha acts as a selector between 0.0 for the base color and 1.0
for the texture color modulated by a constant texture blend color specified with
pfTEnvBlendColor(). The alpha/intensity components are multiplied.
•
PFTE_ADD—The RGB components of the base color are added to the product of the
texture color modulated by the current texture environment blend color. The
alpha/intensity components are multiplied.
Automatic Texture Coordinate Generation
Automatic texture coordinate generation is provided with the pfTexGen state attribute.
pfTexGen closely corresponds to OpenGL’s glTexGen() function. When texture
coordinate generation is enabled, a pfTexGen applied with pfApplyTGen()
automatically generates texture coordinates for all rendered geometry. Texture
coordinates are generated from geometry vertices according to the texture generation
mode set with pfTGenMode(). Available modes and their function are listed in
Table 12-8.
Table 12-8
007-1680-100
Texture Generation Modes
Mode Token
Description
PFTG_OFF
Disables texture coordinate generation.
PFTG_OBJECT_LINEAR
Generates the texture coordinate as the distance from
plane in object space.
PFTG_EYE_LINEAR
Generates the texture coordinate as the distance from
plane in eye space. The plane is transformed by the
inverse of the ModelView matrix when tgen, the
pfTexGen, is applied.
471
12: Graphics State
Table 12-8
Texture Generation Modes (continued)
Mode Token
Description
PFTG_EYE_LINEAR_IDENT
Generates the texture coordinate as the distance from
plane in eye space. The plane is not transformed by the
inverse of the ModelView matrix.
PFTG_SPHERE_MAP
Generates the texture coordinate based on the view
vector reflected about the vertex normal in eye space.
PFTG_OBJECT_DISTANCE_TO_LINE
Sets the texture coordinate as the distance in object
space from the vertex to a line specified with a point and
direction vector through pfTGenPoint.
PFTG_EYE_DISTANCE_TO_LINE
Sets the texture coordinate as the distance in eye space
from the eye to a line specified with pfTGenPoint
through the vertex.
PFTG_REFLECTION_MAP
Sets the texture coordinate to the view vector reflected
about the vertex normal in eye space.
PFTG_NORMAL_MAP
Sets the texture coordinate to the vertex normal in eye
space.
Some modes refer to a plane which is set with pfTGenPlane() and to a line that is
specified as a point and direction with pfTGenPoint(). The default texture generation
mode for all texture coordinates is PFTG_OFF. The function pfGetTGenMode() returns
the mode of the pfTexGen.
See the man page for the OpenGL function glTexGen() for the specific mathematics of
the generation modes for texture coordinates.
Lighting
OpenGL Performer lighting is an extension of graphics library lighting (see glLight() and
related functions in OpenGL). The light embodies the color, position, and type (for
example, infinite or spot) of the light. The light model specifies the environment for
infinite (the default) or local viewing, and two-sided illumination.
The lighting model describes the type of lighting operations to be considered, including
local lighting, two-sided lighting, and light attenuation. The fastest light model is
infinite, single-sided lighting. A pfLightModel state attribute object is created with
472
007-1680-100
Immediate Mode
pfNewLModel(). A light model also allows you to specify ambient light for the scene,
such as might come from the sun with pfLModelAmbient().
The pfLights are created by calling pfNewLight(). A light has color and position. The
light colors are specified with pfLightColor() as follows:
void pfLightColor(pfLightSource* lsource, int which, float r,
float g, float b);
which specifies one of three light colors as follows:
•
PFLT_AMBIENT
•
PFLT_DIFFUSE
•
PFLT_SPECULAR
You to position the light source using pfLightPos():
void pfLightPos(pfLight* light, float x, float y,
float z, float w);
The variable w is the distance between the location in the scene defined by (x, y, z) and
the light source lsource. If w equals 0, lsource is infinitely far away and (x, y, z) defines a
vector pointing from the origin in the direction of lsource; if w equals 1, lsource is located
at the position (x, y, z). The default position is (0, 0, 1, 0), directly overhead and infinitely
far away.
The pfLights are attached to a pfGeoState through the PFSTATE_LIGHTS attribute.
The transformation matrix that is on the matrix stack at the time the light is applied
controls the interpretation of the light source direction:
•
To attach a light to the viewer (like a miner’s head-mounted light), call pfLightOn()
only once with an identity matrix on the stack.
•
To attach a light to the world (like the sun or moon), call pfLightOn() every frame
with only the viewing transformation on the stack.
•
To attach a light to an object (like the headlights of a car), call pfLightOn() every
frame with the combined viewing and modeling transformation on the stack.
The number of lights you can have turned on at any one time is limited by
PF_MAX_LIGHTS, just as is true with the graphics libraries.
007-1680-100
473
12: Graphics State
Note: In previous versions, attenuation was also part of the light model definition. In
OpenGL, attenuation is defined per light . The libpr API for setting it is pfLightAtten().
Note: libpf applications can include light sources in a scene graph with the
pfLightSource node.
Materials
OpenGL Performer materials are an extension of graphics library material (see
glMaterial()). pfMaterials encapsulate the ambient, diffuse, specular, and emissive colors
of an object as well as its shininess and transparency. A pfMaterial is created by calling
pfNewMtl(). As with any of the other attributes, a pfMaterial can be referenced in a
pfGeoState, captured by a display list, or invoked as an immediate mode command.
The pfMaterials, by default, allow object colors to set the ambient and diffuse colors. This
allows the same pfMaterial to be used for objects of different colors, removing the need
for material changes and thus improving performance. This mode can be changed with
pfMtlColorMode(mtl, side, PFMTL_CMODE_*). OpenGL allows front or back materials
to track the current color. If the same material is used for both front and back materials,
there is no difference in functionality.
With the function pfMtlSide() you can specify whether to apply the the material on the
side facing the viewer (PFMTL_FRONT), the side not facing the viewer (PFMTL_BACK),
or both (PFMTL_BOTH). Back-sided lighting will only take affect if there is a two-sided
lighting model active. Two-sided lighting typically has some significant performance
cost.
Object materials only have an effect when lighting is active.
Color Tables
A pfColortable substitutes its own color array for the normal color attribute array
(PFGS_COLOR4) of a pfGeoSet. This allows the same geometry to appear differently in
different views simply by applying a different pfColortable for each view. By leaving the
selection of color tables to the global state, you can use a single call to switch color tables
for an entire scene. In this way, color tables can simulate time-of-day changes, infrared
imaging, psychedelia, and other effects.
474
007-1680-100
Immediate Mode
The pfNewCtab() function creates and returns a handle to a pfColortable. As with other
attributes, you can specify which color table to use in a pfGeoState or you can use
pfApplyCtab() to set the global color table, either in immediate mode or in a display list.
For an applied color table to have effect, color table mode must also be enabled.
Fog
A pfFog is created by calling pfNewFog(). As with any of the other attributes, a pfFog
can be referenced in a pfGeoState, captured by a display list, or invoked as an immediate
mode command. Fog is the atmospheric effect of aerosol water particles that occlude
vision over distance. SGI graphics hardware can simulate this phenomenon in several
different fashions. A fog color is blended with the resultant pixel color based on the range
from the viewpoint and the fog function. pfFog supports several different fogging
methods. Table 12-9 lists the pfFog tokens and their corresponding actions.
Table 12-9
pfFog Tokens
pfFog Token
Action
PFFOG_VTX_LIN
Compute fog linearly at vertices.
PFFOG_VTX_EXP
Compute fog exponentially at vertices (ex).
PFFOG_VTX_EXP2
Compute fog exponentially at vertices (ex squared).
PFFOG_PIX_LIN
Compute fog linearly at pixels.
PFFOG_PIX_EXP
Compute fog exponentially at pixels (ex).
PFFOG_PIX_EXP2
Compute fog exponentially at pixels (ex squared).
PFFOG_PIX_SPLINE
Compute fog using a spline function at pixels.
The pfFogType() function uses these tokens to set the type of fog. A detailed explanation
of fog types is given in the man pages pfFog(3pf) and glFog(3g).
You can set the near and far edges of the fog with pfFogRange(). For exponential fog
functions, the near edge of fog is always zero in eye coordinates. The near edge is where
the onset of fog blending occurs, and the far edge is where all pixels are 100% fog color.
007-1680-100
475
12: Graphics State
The token PFFOG_PIX_SPLINE selects a spline function to be applied when generating
the hardware fog tables. This is further described in the pfFog(3pf) man page. Spline fog
allows you to define an arbitrary fog ramp that can more closely simulate real-world
phenomena like horizon haze.
For best fogging effects, the ratio of the far to the near clipping planes should be
minimized. In general, it is more effective to add a small amount to the near plane than
to reduce the far plane.
Highlights
OpenGL Performer provides a mechanism for highlighting geometry with alternative
rendering styles, useful for debugging and interactivity. A pfHighlight, created with
pfNewHlight(), encapsulates the state elements and modes for these rendering styles. A
pfHighlight can be applied to an individual pfGeoSet with pfGSetHlight() or can be
applied to multiple pfGeoStates through a pfGeoState or pfApplyHlight(). The
highlighting effects are added to the normal rendering phase of the geometry.
pfHighlights make use of special outlining and fill modes and have a concept of a
foreground color and a background color that can both be set with pfHlightColor(). The
available rendering styles can be combined by ORing together tokens for
pfHlightMode() and are described in Table 12-10.
Table 12-10
PFHL_ Mode
Bitmask Token
LINES
pfHlightMode() Tokens
Description
Outlines the triangles in the highlight foreground color according to
pfHlightLineWidth().
476
LINESPAT
LINESPAT2
Outlines triangles with patterned lines in the highlight foreground color or in
two colors using the background color.
FILL
Draws geometry with the highlight foreground color. Combined with
SKIP_BASE, this is a fast highlighting mode.
FILLPAT
FILLPAT2
Draws the highlighted geometry as patterned with one or two colors.
FILLTEX
Draws a highlighting fill pass with a special highlight texture.
LINES_R
FILL_R
Reverses the highlighting foreground and background colors for lines and fill,
respectively.
007-1680-100
Immediate Mode
Table 12-10
pfHlightMode() Tokens (continued)
PFHL_ Mode
Bitmask Token
Description
POINTS
Renders the vertices of the geometry as points according to
pfHlightPntSize().
NORMALS
Displays the normals of the geometry with lines according to
pfHlightNormalLength().
BBOX_LINES
BBOX_FILL
Displays the bounding box of the pfGeoSet as outlines and/or a filled box.
Combined with PFHL_SKIP_BASE, this is a fast highlighting mode.
SKIP_BASE
Causes the normal drawing phase of the pfGeoSet to be skipped. This is
recommended when using PFHL_FILL or PFHL_BBOX_FILL.
For a demonstration of the highlighting styles, see the sample program
/usr/share/Performer/pguide/src/libpr/C/hlcube.c on IRIX and Linux
and %PFROOT%\Src\pguide\libpr\C\hlcube.c on Microsoft Windows.
Graphics Library Matrix Routines
OpenGL Performer provides extensions to the standard graphics library
matrix-manipulation functions. These functions are similar to their graphics library
counterparts, with the exception that they can be placed in OpenGL Performer display
lists. Table 12-11 lists and describes the matrix manipulation routines.
Table 12-11
007-1680-100
Matrix Manipulation Routines
Routines
Action
pfScale()
Concatenate a scaling matrix.
pfTranslate()
Concatenate a translation matrix.
pfRotate()
Concatenate a rotation matrix.
pfPushMatrix()
Push down the matrix stack.
pfPushIdentMatrix()
Push the matrix stack and load an identity matrix on top.
pfPopMatrix()
Pop the matrix stack.
477
12: Graphics State
Table 12-11
Matrix Manipulation Routines (continued)
Routines
Action
pfLoadMatrix()
Add a matrix to the top of the stack.
pfMultMatrix()
Concatenate a matrix.
Sprite Transformations
A sprite is a special transformation used to efficiently render complex geometry with
axial or point symmetry. A classic sprite example is a tree which is rendered as a single,
texture-mapped quadrilateral. The texture image is of a tree and has an alpha component
whose values “etch” the tree shape into the quad. In this case, the sprite transformation
rotates the quad around the tree trunk axis so that it always faces the viewer. Another
example is a puff of smoke which again is a texture-mapped quad but is rotated about a
point to face the viewer so it appears the same from any viewing angle. The pfSprite
transformation mechanism supports both these simple examples as well as more
complicated ones involving arbitrary 3D geometry.
A pfSprite is a structure that is manipulated through a procedural interface. It is different
from attributes like pfTexture and pfMaterial since it affects transformation, rather than
state related to appearance. A pfSprite is activated with pfBeginSprite(). This enables
sprite mode and any pfGeoSet that is drawn before sprite mode is ended with
pfEndSprite() will be transformed by the pfSprite. First, the pfGeoSet is translated to the
location specified with pfPositionSprite(). Then, it is rotated, either about the sprite
position or axis depending on the pfSprite’s configuration. Note that pfBeginSprite(),
pfPositionSprite(), and pfEndSprite() are display listable and this will be captured by
any active pfDispList.
478
007-1680-100
Immediate Mode
A pfSprite’s rotation mode is set by specifying the PFSPRITE_ROT token to
pfSpriteMode(). In all modes, the Y axis of the geometry is rotated to point to the eye
position. Rotation modes are listed below.
Table 12-12
pfSprite Rotation Modes
PFSPRITE_ Rotation Token Rotation Characteristics
AXIAL_ROT
Geometry’s Z axis is rotated about the axis specified with
pfSpriteAxis().
POINT_ROT_EYE
Geometry is rotated about the sprite position with the object
coordinate Z axis constrained to the window coordinate Y axis; that
is, the geometry’s Z axis stays “upright.”
POINT_ROT_WORLD
Geometry is rotated about the sprite position with the object
coordinate Z axis constrained to the sprite axis.
Rather than using the graphics hardware’s matrix stack, pfSprites transform small
pfGeoSets on the CPU for improved performance. However, when a pfGeoSet contains
a certain number of primitives, it becomes more efficient to use the hardware matrix
stack. While this threshold is dependent on the CPU and graphics hardware used, you
may specify it with the PFSPRITE_MATRIX_THRESHOLD token to pfSpriteMode().
The corresponding value is the minimum vertex requirement for hardware matrix
transformation. Any pfGeoSet with fewer vertices will be transformed on the CPU. If you
want a pfSprite to affect non-pfGeoSet geometry, you should set the matrix threshold to
zero so that the pfSprite will always use the matrix stack. When using the matrix stack,
pfBeginSprite() pushes the stack and pfEndSprite() pops the matrix stack so the sprite
transformation is limited in scope.
The pfSprites are dependent on the viewing location and orientation and the current
modeling transformation. You can specify these with calls to pfViewMat() and
pfModelMat(), respectively. Note that libpf-based applications need not call these
routines since libpf does it automatically.
Display Lists
The libpr library supports display lists, which can capture and later execute libpr
graphics commands. pfNewDList() creates and returns a handle to a new pfDispList. A
pfDispList can be selected as the current display list with pfOpenDList(), which puts the
system in display list mode. Any subsequent libpr graphics commands, such as
007-1680-100
479
12: Graphics State
pfTransparency(), pfApplyTex(), or pfDrawGSet() are added to the current display list.
Commands are added until pfCloseDList() returns the system to immediate mode. It is
not valid to have multiple pfDispLists open at a given time but a pfDispList may be
reopened, in which case, commands are appended to the end of the list.
Once a display list is constructed, it can be executed by calling pfDrawDList(), which
traverses the list and sends commands down the Geometry Pipeline.
The pfDispLists are designed for multiprocessing, where one process builds a display list
of the visible scene and another process draws it. The function pfResetDList() facilitates
this by making pfDispLists reusable. Commands added to a reset display list overwrite
any previously entered commands. A display list is typically reset at the beginning of a
frame and then filled with the visible scene.
The pfDispLists support concurrent multiprocessing, where the producer and consumer
processes simultaneously write and read the display list. The PFDL_RING argument to
pfNewDList() creates a ring buffer or FIFO-type display list. pfDispLists automatically
ensure ring buffer consistency by providing synchronization and mutual exclusion to
processes on ring buffer full or empty conditions.
For more information and the application of display lists, see Chapter 15, “ClipTextures.”
Combining Display Lists
The contents of one pfDispList may be appended to a second pfDispList by using the
function, pfAppendDList(). All pfDispList elements in src are appended to the
pfDispList dlist.
Alternately, you can append the contents of one pfDispList to a second pfDispList by
using the function pfDispList::append(). All pfDispList elements in src are appended to
the pfDispList on which the append method is invoked.
State Management
A pfState is a structure that represents the entire libpr graphics state. A pfState
maintains a stack of graphics states that can be pushed and popped to save and restore
the state. The top of the stack describes the current graphics state of a window as it is
known to OpenGL Performer.
480
007-1680-100
Immediate Mode
The pfInitState() function initializes internal libpr state structures and should be called
at the beginning of an application before any pfStates are created. Multiprocessing
applications should pass a usinit() semaphore arena pointer to pfInitState(), such as
pfGetSemaArena(), so OpenGL Performer can safely manage state between processes.
pfNewState() creates and returns a handle to a new pfState, which is typically used to
define the state of a single window. If using pfWindows, discussed in Chapter 16,
“Windows,” a pfState is automatically created for the pfWindow when the window is
opened and the current pfState is switched when the current pfWindow changes.
pfSelectState() can be used to efficiently switch a different, complete pfState.
pfLoadState() forces the full application of a pfState.
Pushing and Popping State
The pfPushState() function pushes the state stack of the currently active pfState,
duplicating the top state. Subsequent modifications of the state through libpr routines
are recorded in the top of the stack. Consequently, a call to pfPopState() restores the state
elements that were modified after pfPushState().
The code fragment in Example 12-2 illustrates how to push and pop state.
Example 12-2
Pushing and Popping Graphics State
/* set state to transparency=off and texture=brickTex */
pfTransparency(PFTR_OFF);
pfApplyTex(brickTex);
/* ... draw geometry here using original state ... */
/* save old state. establish new state */
pfPushState();
pfTransparency(PFTR_ON);
pfApplyTex(woodTex);
/* ... draw geometry here using new state ...*/
/* restore state to transparency=off and texture=brickTex */
pfPopState();
007-1680-100
481
12: Graphics State
State Override
The pfOverride() function implements a global override feature for libpr graphics state
and attributes. pfOverride() takes a mask that indicates which state elements to affect
and a value specifying whether the elements should be overridden. The mask is a bitwise
OR of the state tokens listed previously.
The values of the state elements at the time of overriding become fixed and cannot be
changed until pfOverride() is called again with a value of zero to release the state
elements.
The code fragment in Example 12-3 illustrates the use of pfOverride().
Example 12-3
Using pfOverride()
pfTransparency(PFTR_OFF);
pfApplyTex(brickTex);
/*
* Transparency will be disabled and only the brick texture
* will be applied to subsequent geometry.
*/
pfOverride(PFSTATE_TRANSPARENCY | PFSTATE_TEXTURE, 1);
/* Draw geometry */
/* Transparency and texture can now be changed */
pfOverride(PFSTATE_TRANSPARENCY | PFSTATE_TEXTURE, 0);
pfGeoState
A pfGeoState encapsulates all the rendering modes, values, and attributes managed by
libpr. See “Rendering Modes” on page 453, “Rendering Values” on page 458, and
“Rendering Attributes” on page 459 for more information. pfGeoStates provide a
mechanism for combining state into logical units and define the appearance of geometry.
For example, you can set a brick-like texture and a reddish-orange material on a pfGeoSet
and use it when drawing brick buildings.
You can specify texture matricies on pfGeoSets.
482
007-1680-100
Immediate Mode
Local and Global State
There are two levels of rendering state: local and global. A record of both is kept in the
current pfState. The local state is that defined by the settings of the current pfGeoState.
The rendering state and attributes of a pfGeoState can be either locally set or globally
inherited. If all state elements are set locally, a pfGeoState becomes a full graphics
context—that is, all state is then defined at the pfGeoState level. Global state elements are
set with libpr immediate-mode routines like pfEnable(), pfApplyTex(), pfDecal(), or
pfTransparency() or by drawing a pfDispList containing these commands with
pfDrawDList(). Local state elements for subsequent pfGeoSets are set by applying a
pfGeoState with pfApplyGState() (note that pfDrawGSet() automatically calls
pfApplyGState() if the pfGeoSet has an attached pfGeoState). The state elements applied
by a pfGeoState are those modes, enables, and attributes that are explicitly set on the
pfGeoState. Those settings revert back to the pfState settings for the next call to
pfApplyGState(). A pfGeoState can be explicitly loaded into a pfState to affect future
pfGeoStates with pfLoadGState().
Note: By default, all state elements are inherited from the global state. Inherited state
elements are evaluated faster than values that have been explicitly set.
While it can be useful to have all state defined at the pfGeoState level, it usually makes
sense to inherit most state from global default values and then explicitly set only those
state elements that are expected to change often.
Examples of useful global defaults are lighting model, lights, texture environment, and
fog. Highly variable state is likely to be limited to a small set such as textures, materials,
and transparency. For example, if the majority of your database is lighted, simply
configure and enable lighting at the beginning of your application. All pfGeoStates will
be lighted, except the ones for which you explicitly disable lighting. Then, attach
different pfMaterials and pfTextures to pfGeoStates to define specific state combinations.
Note: Use caution when enabling modes in the global state. These modes may have cost
even when they have no visible effect. Therefore, geometry that cannot use these modes
should have a pfGeoState that explicitly disables the mode. Modes that require special
care include the texturing enable and transparency.
You specify that a pfGeoState should inherit state elements from the global default with
pfGStateInherit(gstate, mask). mask is a bitmask of tokens that indicates which state
007-1680-100
483
12: Graphics State
elements to inherit. These tokens are listed in the “Rendering Modes” on page 453,
“Rendering Values” on page 458, and “Rendering Attributes” on page 459 sections of
this chapter. For example, PFSTATE_ENLIGHTING | PFSTATE_ENTEXTURE makes
gstate inherit the enable modes for lighting and texturing.
A state element ceases to be inherited when it is set in a pfGeoState. Rendering modes,
values, and attributes are set with pfGStateMode(), pfGStateVal(), and pfGStateAttr(),
respectively. For example, to specify that gstate is transparent and textured with treeTex,
use the following:
pfGStateMode(gstate, PFSTATE_TRANSPARENCY, PFTR_ON);
pfGStateAttr(gstate, PFSTATE_TEXTURE, treeTex);
Applying pfGeoStates
Use pfApplyGState() to apply the state encapsulated by a pfGeoState to the Geometry
Pipeline. The effect of applying a pfGeoState is similar to applying each state element
individually. For example, if you set a pfTexture and enable a decal mode on a
pfGeoState, applying it essentially calls pfApplyTex() and pfDecal(). If in display-list
mode, pfApplyGState() is captured by the current display list.
State is (logically) pushed before and popped after pfGeoStates are applied so that
pfGeoStates do not inherit state from each other. This is a very powerful and convenient
characteristic since, as a result, pfGeoStates are order-independent, and you do not have
to worry about one pfGeoState corrupting another. The code fragment in Example 12-4
illustrates how pfGeoStates inherit state.
Example 12-4
Inheriting State
/* gstateA should be textured */
pfGStateMode(gstateA, PFSTATE_ENTEXTURE, PF_ON);
/* gstateB inherits the global texture enable mode */
pfGStateInherit(gstateB, PFSTATE_ENTEXTURE);
/* Texturing is disabled as the global default */
pfDisable(PFEN_TEXTURE);
/* Texturing is enabled when gstateA is applied */
pfApplyGState(gstateA);
/* Draw geometry that will be textured */
/* The global texture enable mode of OFF is restored
484
007-1680-100
Immediate Mode
so that gstateB is NOT textured. */
pfApplyGState(gstateB);
/* Draw geometry that will not be textured */
The actual pfGeoState pop is a "lazy" pop that does not happen unless a subsequent
pfGeoState requires the global state to be restored. This means that the actual state
between pfGeoStates is not necessarily the global state. If a return to global state is
required, call pfFlushState() to restore the global state. Any modification to the global
state made using libpr functions—pfTransparency(), pfDecal(), and so on—becomes
the default global state.
For best performance, set as little local pfGeoState state as possible. You can accomplish
this by setting global defaults that satisfy the majority of the requirements of the
pfGeoStates being drawn. By default, all pfGeoState state is inherited from the global
default.
pfGeoSets and pfGeoStates
There is a special relationship between pfGeoSets and pfGeoStates. Together they
completely define both geometry and graphics state. You can attach a pfGeoState to a
pfGeoSet with pfGSetGState() to specify the appearance of geometry. Whenever the
pfGeoSet is drawn with pfDrawGSet(), the attached pfGeoState is first applied using
pfApplyGState(). If a pfGeoSet does not have a pfGeoState, its state description is
considered undefined. To inherit all values from the global pfState, a pfGeoSet should
have a pfGeoState with all values set to inherit, which is the default.
This combination of routines allows the application to combine geometry and state in
high-performance units that are unaffected by rendering order. To further increase
performance, sharing pfGeoStates among pfGeoSets is encouraged.
Table 12-13 lists and describes the pfGeoState routines.
Table 12-13
007-1680-100
pfGeoState Routines
Routiine
Description
pfNewGState()
Create a new pfGeoState.
pfCopy()
Make a copy of the pfGeoState.
pfDelete()
Delete the pfGeoState.
pfGStateMode()
Set a specific state mode.
485
12: Graphics State
Table 12-13
pfGeoState Routines (continued)
Routiine
Description
pfGStateVal()
Set a specific state value.
pfGStateAttr()
Set a specific state attribute.
pfGStateInherit()
Specify which state elements are inherited from the global state.
pfApplyGState()
Apply pfGeoState’s non-inherited state elements to graphics.
pfLoadGState()
Load pfGeoState’s settings into the pfState, inherited by future
pfGeoStates.
pfGetCurGState()
Return the current pfGeoState in effect.
pfGStateFuncs()
Assign pre/post callbacks to pfGeoState.
pfApplyGStateTable()
Specify a able of pfGeoStates used for indexing.
Figure 12-1 diagrams the conceptual structure of a pfGeoState.
486
007-1680-100
Immediate Mode
t
igh
pfL del
Mo
a
pfM
two
loc -side
al/i fla
nfin g
am ite fla
b
atte ient g
nua
tion
on/
pf
of
am f tag
bie
col nt
pos or
dire ition
c
spr tion
e
exp ad
one
nt
p
mo
des
tat
Lig
e
ues
ligh
tm
ode
l
ligh
ts [
fro
8]
nt m
ate
rial
tex
tur
e
ee
nvi
rom
fog
ial
ent
or t
tex
ter
alp
am ha
bi
diff ent
em use
is
shi sion
nin
spe ess
cul
ar
rial
km
col
si
iz
iz
i
com ze com e com ze com e
po
p
p
p
couonen couonent couonen cou nent
n t nt
n t nt
ima t ima
ima t ima
ge
ge
ge
ge
mo
mo
mo
mo
d
d
d
d
det esdeta esdet esdeta es
a
ail
tex il text il tex il text
ure
tur
tur
ure
e
e
a
pfM
ate
bac
tur
e
ure ure ure tur
ext fText fText pfTex
ps
ps
ial
alp
am ha
bi
diff ent
em use
is
shi sion
ni
spe ness
cul
ar
val
ht
tex
psfT
oS
fGe
ter
abl
e
mo
tur
del
eg
pfL
ene
v
Env Env EnvxEn
rati
Poi
x
tex tex texpfTteextfTexpfTexpfTe
tur tu tur p ure
e in re i e in in
d
ble bdleex nbdleex dbelexn ex
nd nd nd d
m
mo mo mo od
des de des es
s
on
ntS
tate
og
col
pfF
or
sta
rt
end
le
e
ure ure ure tur
ext fText fText pfTex
T
f
p
ps
ps
s
si
iz
iz
i
com ze com e com ze com e
po
p
p
p
couonen couonent couonen cou nent
n t nt
n t nt
ima t ima
ima t ima
ge
ge
ge
ge
mo
mo
mo
mo
d
d
d
d
det esdeta esdet esdeta es
a
ail
tex il text il tex il text
ure
tur
tur
ure
e
e
Figure 12-1
007-1680-100
tab
en
eGxeGn
T
f
p ex n
pfeTxGen
T
f
p exGe
pfT
rtab
lo
fCo
p
le a
d
tab dress
le s
ize
pfGeoState Structure
487
12: Graphics State
Multitexture Support in pfGeoState
Some graphic hardware supports the use of multiple texture maps on a single polygon.
These multiple texture maps are blended together according to a collection of texture
environments. Figure 12-2 demonstrates the OpenGL definition for generating the color
of a multitextured pixel. The figure assumes that the hardware has four texture units and
so each pixel can receive contribution from four texture maps.
:
Shaded color
Texture 0
Texture 1
Texture 2
Texture 3
Figure 12-2
TexEnv 0
TexEnv 1
TexEnv 2
TexEnv 3
Final Pixel Color
Generating the Color of a Multitextured Pixel
In the figure, the shaded and un-textured color of a pixel enters the first texture blending
unit together with the texture color computed by the first texture unit. The texture
environment marked TexEnv 0 determines the math operation between the two. The
output color of this operation feeds the second texture blending unit together with the
texture color computed by the second texture unit. This process continues four times
until the final color of the pixel is generated.
The pfGeoState class allows specifying multiple texture maps on a single pfGeoSet. All
these texture maps will be applied when the pfGeoSet is applied (providing that the
graphic hardware has enough texture mapping units). pfGeoState also allows specifying
multiple pfTexEnv, pfTexGen, pfTexMat, and pfTexLOD objects—one for each pfTexture.
488
007-1680-100
Immediate Mode
The following code fragment shows how to add multiple textures to a pfGeoState:
{
pfGeoState *gstate;
pfTexture *tex;
pfTexEnv
*tev;
gstate = pfNewGState (pfGetSharedArena());
for (i = 0 ; i < PF_MAX_TEXTURES ; i ++)
{
/* Load texture # i from a file */
tex = pfNewTex (pfGetSharedArena());
pfLoadTexFile (tex, texture_file_name[i]);
tev = pfNewTEnv (pfGetSharedArena());
/* Enable texture unit # i on the pfGeoState. */
pfGStateMultiMode (gstate, PFSTATE_ENTEXTURE, i, 1);
/* Attach texture for texture unit # i */
pfGStateMultiAttr (gstate, PFSTATE_TEXTURE, i, tex);
/* Attach texture environment for texture unit # i */
pfGStateMultiAttr (gstate, PFSTATE_TEXENV, i, tev);
}
}
Notes:
007-1680-100
•
pfGeoState recognizes texture units starting at the first array (index of 0) and ending
immediately before the first disabled texture unit. For example, enabling texture
units 0, 1, and 3 is equivalent to enabling only texture units 0 and 1.
•
pfGeoState can inherit all or none of the texture units. It is enough to specify one
texture unit in order to avoid inheriting any other texture unit. In order to inherit all
texture units, one must specify no texture units on the pfGeoState.
•
For every texture unit enabled, the application must provide texture coordinates.
Neither OpenGL Performer nor OpenGL will share texture coordinates between
texture units. There are two ways to set texture coordinates:
•
Specifying a pfTexGen for a texture unit
•
Specifying a texture-coordinate attribute array for a texture unit on a pfGeoSet
See section “Attributes” in Chapter 8.
489
Chapter 13
13. Shaders
Recent graphics hardware allows you to write vertex and fragment programs to replace
the corresponding fixed functionality of earlier systems. An independently compiliable
unit of such programs is referred to as a shader. OpenGL Performer supports the
OpenGL Shading Language (GLSL), which provides a platform-independent interface
for such programs.
This chapter describes the OpenGL Performer interface to GLSL. This implementation is
based on two new classes, which are described in the following sections:
•
“The pfShaderProgram Class” on page 491
•
“The pfShaderObject Class” on page 497
•
“Example Code” on page 500
Note: In Perfly, you can enable/disable shader programs by using a keyboard entry of a.
The pfShaderProgram Class
The pfShaderProgram class encapsulates the functionality associated with OpenGL
shader programs. A pfShaderProgram is a comprised of a collection of pfShaderObjects
and an collection of uniform variables that are an opaque type and can only be accessed
through an index from the pfShaderProgram interface.
This section covers the following topics:
007-1680-100
•
“Allocating Memory for a Shader Program”
•
“Creating a Shader Program”
•
“Applying Shader Programs”
491
13: Shaders
Allocating Memory for a Shader Program
The function pfNewShaderObject() creates and returns a handle to a pfShaderProgram.
The value arena specifies a malloc() arena out of which the pfShaderProgram is allocated
or NULL for allocation off the process heap. You can delete pfShaderPrograms with
pfDelete().
The function new(arena) allocates a pfShaderProgram from the specified memory arena,
or from the heap if arena is NULL. The function allocates a pfShaderProgram from the
default memory arena (see the pfGetSharedArena man page). Like other pfObjects,
pfShaderPrograms cannot be automatically created statically on the stack or in arrays.
Delete pfShaderPrograms with pfDelete() rather than with the delete operator.
The function pfGetShaderObjectClassType() returns the pfType* for the class
pfShaderProgram. The pfType* returned by pfGetShaderObjectClassType() is the
same as the pfType* returned by invoking pfGetType(), the virtual function getType()
on any instance of class pfShaderProgram. Because OpenGL Performer allows
subclassing of built-in types when decisions are made based on the type of an object, use
pfIsOfType() the member function isOfType() to test if an object is of a type derived
from an OpenGL Performer type rather than to test for strict equality of the pfType*s.
In order to use a pfShaderProgram as a piece of state, you must specify it as an attribute
for a pfGeoState and enable that mode, as shown in the following code:
pfGeoState *gState = pfNewGState(arena);
pfShaderProgram *sProg = pfNewSProg(arena);
pfGStateMode(gState, PFSTATE_ENSHADPROG, PF_ON);
pfGStateAttr(gState, PFSTATE_SHADPROG, sProg);
492
007-1680-100
The pfShaderProgram Class
Creating a Shader Program
Creating a valid pfShaderProgram involves the following steps:
•
Adding a set of pfShaderObjects to the shader program
•
Specifying uniform variables (optional)
•
Clamping uniform variables (optional)
•
Normalizing uniform variables (optional)
Adding pfShaderObjects to a Shader program
For shader programs, you can add, replace, and delete pfShaderObjects with the
following functions, respectively:
pfSProgAddShader()
pfSProgReplaceShader()
pfSProgRemoveShader()
In addition to the methods for adding/deleting/replacing shader objects from a shader
program, you can query an existing pfShaderProgram to determine the number of
associated shader objects with pfGetSProgNumShaders(). To retreive a pointer to the ith
shader object, use pfGetSProgShader().
Specifying Uniform Variables (optional)
In addition to specifying a set of pfShaderObjects for a pfShaderProgram, you can specify
a set of uniform variables for the program. These uniforms, which are not necessarily the
same as those set internally by OpenGL, can then be referenced by index.
A uniform variable is comprised of the following:
007-1680-100
•
A name (stored as a GLcharARB*)
•
A type
•
A value indicating the number of variables of that type being stored in the uniform
•
A flag indicating if the value is to be clamped to a minimum and/or maximum
value (or not at all)
493
13: Shaders
•
A flag indicating if the value should be normalized (for those types which can be
normalized)
•
Pointers to the current value (as well as the minimum and maximum values for this
uniform variable if, indeed, it is clamped)
Table 13-1 shows the uniform variable types.
Table 13-1
494
Uniform Variable Types
Type
Description
PFUNI_FLOAT1
Single UGLfloat
PFUNI_FLOAT2
Array of two GLfloats
PFUNI_FLOAT3
Array of three GLfloats
PFUNI_FLOAT4
Array of four GLfloats
PFUNI_INT1
Single GLint
PFUNI_INT2
Array of two GLints
PFUNI_INT3
Array of three GLints
PFUNI_INT4
Array of four GLints
PFUNI_BOOL1
Single GLint specifying boolean value
PFUNI_BOOL2
Array of two GLints specifying boolean values
PFUNI_BOOL3
Array of three GLints specifying boolean values
PFUNI_BOOL4
Array of four GLints specifying boolean values
PFUNI_MAT2
Four GLfloats specifying 2x2 matrix
PFUNI_MAT3
Nine GLfloats specifying 3x3 matrix
PFUNI_MAT4
Sixteen GLfloats specifying 4x4 matrix
PFUNI_SAMP1D
Single GLint specifying which texture unit to query for this
sampler
PFUNI_SAMP2D
Single GLint specifying which texture unit to query for this
sampler
007-1680-100
The pfShaderProgram Class
Table 13-1
Uniform Variable Types (continued)
Type
Description
PFUNI_SAMP3D
Single GLint specifying which texture unit to query for this
sampler
PFUNI_SAMPCUBE
Single GLint specifying which texture unit to query for this
sampler
PFUNI_SAMP1DSHADOW
Single GLint specifying which texture unit to query for this
sampler
PFUNI_SAMP2DSHADOW
Single GLint specifying which texture unit to query for this
sampler
Adding Uniform Variables to Shader Programs
In order to add a uniform variable to a pfShaderProgram, use pfSprogAddUniform(),
whose parameters follow:
name
Specifies the name for the uniform variable.
uniType
Specifies one of the types listed in Table 13-1.
size
Indicates how many variables of this type will be found in the fourth
and final parameter data.
Internally, OpenGL Performer will make a copy of this data if it is not a pointer to a
pfMemory; if it is, then the reference count for this piece of memory will get incremented
and no copy will be performed.
For example, the following code adds a uniform variable called scaleFactor, which is
a 4-byte float (size of a GLfloat) set to 0.5:
int uniformIndex;
GLfloat scale = 0.5f;
pfShaderProgram *sProg = pfNewSprog(arena);
uniformIndex = pfSProgAddUniform(sProg, "scaleFactor", PFUNI_FLOAT1, 1,
&scale);
007-1680-100
495
13: Shaders
Clamping Uniform Variables (optional)
You can set a flag to indicate whether or not a uniform variable should be clamped and
if it should be normalized. A uniform variable can be clamped by using
pfSProgClampMode() with any one of the following parameters:
•
PFUNI_CLAMP_NONE
•
PFUNI_CLAMP_MIN
•
PFUNI_CLAMP_MAX
•
PFUNI_CLAMP_ALL
The default value is PFUNI_CLAMP_NONE. In order to set the clamp value, call
pfSProgUniformMin() or pfSProgUniformMax(). The following code shows an
example:
GLfloat minValue = 0.0f;
GLfloat maxValue = 1.0f;
pfSProgUniformMin(sProg, uniformIndex, &minValue);
pfSProgUniformMax(sProg, uniformIndex, &maxValue);
In order to query the minimum and maximum values, use pfGetSProgUniformMin()
and pfGetSProgUniformMax(). In order to determine if the value is being clamped used,
you can use pfGetSProgClampMode() and check the return value.
Normalizing Uniform Variables (optional)
For uniform variables of type PFUNI_FLOAT2, PFUNI_FLOAT3, or PFUNI_FLOAT4, it is
also possible to normalize the values such that they correspond to vectors of size 1.0. In
order to do this, you must use pfSProgNormalizeFlag() and this flag may also be queried
for a given uniform variable with pfGetSProgNormalizeFlag(). By default, uniform
variables are not normalized. If this flag is set on a uniform variable of a type other than
one that supports this feature, the flag will be ignored and a warning will be issued at run
time.
Applying Shader Programs
You can apply a pfShaderProgram using pfShaderProgramApply(), but only in the draw
process. When this operation is performed, OpenGL Performer will determine if the
shader program needs to be recompiled and perform that operation if required. It is also
496
007-1680-100
The pfShaderObject Class
possible to force compilation to occur by calling pfSProgLink(). This method will return
the following:
Return Value
Meaning
0
The complilation was successful.
–1
The associated OpenGL handle is NULL or some other event occurred.
n
The link process failed. The positive integer n indicates how many
pfShaderObjects did not compile.
In the case where one would like to force OpenGL Performer to relink a
pfShaderProgram once, call pfSProgForceRelink().
One can verify if a given pfShaderProgram is in a state where it is ready to be applied by
calling pfSProgValidate(), which will return 1 if the program can be applied given the
current state or 0, otherwise.
The OpenGL handle associated with a given pfShaderProgram can be retrieved using
pfGetSProgHandle().
The pfShaderObject Class
The pfShaderObject is a class that encapsulates the functionality associated with either
vertex or fragment programs used by the OpenGL Shading Language.
A pfShaderObject is represented by a string containing the source code and a shader
type. A collection of pfShaderObjects can be assembled to form a valid
pfShaderProgram, which can then be used as a piece of state used by pfGeoState with the
PFSTATE_SHADPROG attribute.
This section describes the following topics:
007-1680-100
•
“Creating New Shader Objects”
•
“Specifying Shader Objects”
•
“Specifying the Object Type”
•
“Compiling Shader Objects”
497
13: Shaders
Creating New Shader Objects
The function pfNewShaderObject() creates a new shader object and returns a handle to
a pfShaderObject. The value arena specifies a malloc() arena out of which the
pfShaderObject is allocated or NULL for allocation off the process heap. You can delete
pfShaderObjects with pfDelete().
The function call new(arena) allocates a pfShaderObject from the specified memory arena
or from the heap if arena is NULL. The function allocates a pfShaderObject from the
default memory arena (see the pfGetSharedArena man page). Like other pfObjects,
pfShaderObjects cannot be automatically created statically on the stack or in arrays.
Delete pfShaderObjects with pfDelete() rather than the delete operator.
The function pfGetShaderObjectClassType() returns the pfType* for the class
pfShaderObject. The pfType* returned by pfGetShaderObjectClassType() is the same
as the pfType* returned by invoking pfGetType(), the virtual function getType() on any
instance of class pfShaderObject. Because OpenGL Performer allows subclassing of
built-in types when decisions are made based on the type of an object, use pfIsOfType(),
the member function isOfType(), to test if an object is of a type derived from an
OpenGL Performer type rather than to test for strict equality of the pfType*s.
Specifying Shader Objects
A pfShaderObject can be specified either by loading the source explicitly or by simply
specifying the filename parameter with pfShaderObjectName(). The location for
shader objects specified by filename corresponds to the semantics of pfFindFile() and,
hence, the PFPATH environment variable can be used to specify the location of source
files. In order to retreive the name of the current shader source, you can use
pfGetShaderObjectName(). If the return srting is NULL, it means that the source is
inlined, not loaded from an external file. If the shader source is loaded from a file, you
must also call pfShaderObjectLoad() in order to load the source code into the shader
object.
The source code for the shader object can also be specified explicitly with
pfShaderObjectSource(). The corresponding get method is
pfGetShaderObjectSource().
498
007-1680-100
The pfShaderObject Class
Specifying the Object Type
In addition to setting the source in the form of either a filename or an ASCII string, you
must also specify the shader type for a pfShaderObject. By default, this is set to –1
(invalid) and it must be set to either PFSHAD_FRAGMENT_SHADER or
PFSHD_VERTEX_SHADER with pfShaderObjectShaderType(). You can also retrieve the
shader type for a given pfShaderObject with pfGetShaderObjectShaderType().
If either the source code for the shader object or the type for the shader object have
changed, the compilation status for the shader object will change. One can determine the
necessity for recompiling a given pfShaderObject by calling
pfGetShaderObjectCompileStatus(), which will return 1 if recompilation is required
and 0, otherwise. This is used internally to determine if a pfShaderProgram needs to be
re-linked and, hence, should not normally be called from a user-specified program.
Compiling Shader Objects
A pfShaderObject may be compiled with pfShaderObjectCompile(). The log for the
compilation process will be stored in the log parameter. If successful, the compilation
will return 1 and 0, otherwise.
Once a pfShaderObject has been bound to the current graphics context, you can retrieve
its GL handle with pfGetShaderObjectHandle(). The handle is created as needed during
the compilation process. If the pfShaderObject has not yet been compiled, then the
handle will be NULL.
007-1680-100
499
13: Shaders
Example Code
The following example illustrates the use of the pfShaderProgram and pfShaderObject
classes.
#include
#include
#include
#include
#include
#include
<Performer/pr/pfGeoState.h>
<Performer/pr/pfGeoArray.h>
<Performer/pr/pfTexture.h>
<Performer/pf/pfGeode.h>
<Performer/pr/pfShaderObject.h>
<Performer/pr/pfShaderProgram.h>
#include <Performer/pfdu.h>
int main(int argc,char *argv[])
{
int i;
pfInit();
pfdInitConverter("pfb");
pfConfig();
char *vertexShaderSource =
"varying vec2 tc;\n\n"
"void main() {\n"
" tc = gl_MultiTexCoord0.xy;\n"
" gl_Position = gl_ModelViewProjectionMatrix * gl_Vertex;\n"
"}\n";
FILE *fp = fopen("multiTex.frag", "w");
if(fp) {
for(i=1; i<argc; i++)
fprintf(fp,"uniform sampler2D texture_%d;\n",i-1);
fprintf(fp,"varying vec2 tc;\n");
fprintf(fp,
"void main()\n"
"{\n");
float stepSize = 1.0/(argc-1);
fprintf(fp,"
float stepSize = %g;\n",stepSize);
for(i=0; i<argc-1; i++) {
fprintf(fp,
500
007-1680-100
Example Code
"
"
if(tc.x >= stepSize*%.1f && tc.x <= stepSize*%.1f)\n"
gl_FragColor = texture2D(texture_%d, vec2(tc.x*%.1f,
tc.y));\n",
float(i), float(i+1), i, float(argc-1));
if((i+1) != (argc-1))
fprintf(fp," else ");
}
fprintf(fp,"}\n");
fclose(fp);
} else {
pfNotify(PFNFY_FATAL,PFNFY_PRINT,
"Unable to open multiTex.frag for writing.");
pfExit();
return 1;
}
pfShaderObject *soV = new pfShaderObject();
soV->setShaderType(PFSHD_VERTEX_SHADER);
soV->setSource(vertexShaderSource);
pfShaderObject *so = new pfShaderObject();
so->setShaderType(PFSHD_FRAGMENT_SHADER);
so->setName("multiTex.frag");
pfShaderProgram *sp = new pfShaderProgram();
sp->addShader(so);
sp->addShader(soV);
pfGeoState *gs = new pfGeoState();
gs->setMode(PFSTATE_ENSHADPROG, PF_ON);
gs->setAttr(PFSTATE_SHADPROG, sp);
for(i=1; i<argc; i++) {
pfTexture *tex = new pfTexture();
tex->setName(argv[i]);
int uniValue = i-1;
char name[64];
sprintf(name,"texture_%d",i-1);
sp->addUniform(name, PFUNI_SAMP2D, 1, &uniValue);
007-1680-100
501
13: Shaders
pfNotify(PFNFY_NOTICE,PFNFY_PRINT,"Setting texture %d to %s, named
%s",
i-1,argv[i],name);
gs->setMultiAttr(PFSTATE_TEXTURE, i-1, tex);
}
pfVec3 *verts = (pfVec3 *)pfMalloc(sizeof(pfVec3) * 4);
PFSET_VEC3(verts[0],
PFSET_VEC3(verts[1],
PFSET_VEC3(verts[2],
PFSET_VEC3(verts[3],
0.f, 0.f, 0.f
(argc-1)*1.f,
(argc-1)*1.f,
0.f, 0.f, 1.f
);
0.f, 0.f );
0.f, 1.f );
);
pfVec2 *tCoords = (pfVec2 *)pfMalloc(sizeof(pfVec2) * 4);
PFSET_VEC2(tCoords[0], 0.f, 0.f);
PFSET_VEC2(tCoords[1], 1.f, 0.f);
PFSET_VEC2(tCoords[2], 1.f, 1.f);
PFSET_VEC2(tCoords[3], 0.f, 1.f);
int lengths[1] = { 4 };
pfGeoArray *ga = new pfGeoArray();
ga->setNumPrims(1);
ga->setPrimType(PFGS_TRIFANS);
ga->setPrimLengths(lengths);
ga->setAttr(PFGA_COORD_ARRAY, 3, GL_FLOAT, 0, verts);
// we can use the same set of tex coords for all ...
ga->setAttr(PFGA_TEX_ARRAY, 2, GL_FLOAT, 0, tCoords);
ga->setGState(gs);
pfGeode *geode = new pfGeode();
geode->addGSet(ga);
pfdStoreFile(geode,
getenv("OUTFILE")?getenv("OUTFILE"):"multiTexShader.pfb");
pfExit();
return 0;
}
502
007-1680-100
Chapter 14
14. Using Scalable Graphics Hardware
Scalable graphics hardware provides nearly perfect scaling of both geometry rate and fill
rate on some applications. This chapter describes how you use OpenGL Performer in
conjunction with an SGI Video Digital Multiplexer (DPLEX), an SGI Scalable Graphics
Compositor, and graphics processing units (GPUs). The corresponding sections are the
following:
•
“Using OpenGL Performer with a DPLEX” on page 503
•
“Using OpenGL Performer with an SGI Scalable Graphics Compositor” on page 517
•
“Using OpenGL Performer with GPUs” on page 532
Using OpenGL Performer with a DPLEX
A DPLEX is an optional daughtercard that permits multiple graphics hardware pipelines
to work simultaneiously on a single visual application. DPLEX hardware is available on
Silicon Graphics Onyx2, SGI Onyx 3000, and SGI Onyx 300 systems. For an overview of
the DPLEX hardware, see the document Onyx2 DPLEX Option Hardware User’s Guide.
OpenGL Performer taps the power of a DPLEX by using hyperpipes. The following
sections describe how to use hyperpipes:
007-1680-100
•
“Hyperpipe Concepts” on page 504
•
“Configuring Hyperpipes” on page 504
•
“Configuring pfPipeWindows and pfChannels” on page 511
•
“Programming with Hyperpipes” on page 515
503
14: Using Scalable Graphics Hardware
Hyperpipe Concepts
A pfHyperpipe is a combination of pfPipes or pfMultipipes; there is one pfPipe for each
graphics pipe in a DPLEX ring or chain. A DPLEX ring or chain is a collection of
interconnected graphic boards.
A key concept with hyperpipes is that of temporal decomposition. Think of a rendered
sequence as a 3D data set with time being the third axis. With temporal decomposition,
the dataset is subdivided along the time axis and distributed across, in this case, each of
the graphic pipes in the hyperpipe group.
Temporal decomposition is different from spatial decomposition, in which the dataset is
subdivided along the X axis, Y axis, or both X and Y axes.
Configuring Hyperpipes
It is the responsibility of the application to establish the hyperpipe group configuration
for OpenGL Performer. There are two steps in the configuration process:
1.
Establish the number of graphic pipes (or pfPipes because there is a one-to-one
correspondence) in each hyperpipe group.
2. Map the pfPipes to specific graphic pipes.
Establishing the Number of Graphic Pipes
Use the argument in the pfHyperpipe() function to establish the number of graphic pipes
in the hyperpipe group, for example:
pfHyperpipe(2);
pfConfig();
In this example, two pfPipes combine to create the pfHyperpipe, as shown in Figure 14-1.
504
007-1680-100
Using OpenGL Performer with a DPLEX
h
pfC
an
ne
l
p
p
fHy
erp
pfP
pfP
ipe
ipe
ipe
pfPipeWindow
pfPipeWindow
Figure 14-1
pfPipes Creating pfHyperpipes
Like the pfMultipipe() function, pfHyperpipe() must be invoked prior to configuring
the pfPipes using pfConfig() and after the call to pfInit().
The number of pipes is used by pfConfig() to associate the configured pfPipes. The
pfHyperpipe() function can be invoked multiple times to construct multiple hyperpipe
groups, as shown in Figure 14-2.
007-1680-100
505
14: Using Scalable Graphics Hardware
pfC
pfC
h
an
l
ne
p
p
fHy
erp
pfP
pfP
ha
e
nn
ipe
l
p
pfP
p
fHy
erp
pfP
ipe
ipe
ipe
ipe
ipe
pfPipeWindow
pfPipeWindow
pfPipeWindow
pfPipeWindow
pfH
e
yp
pfP
pfP
rpi
pe
pfP
ipe
ipe
ipe
pfPipeWindow
pfPipeWindow
pfPipeWindow
Figure 14-2
Multiple Hyperpipes
Additionally, the pfHyperpipe() function can be combined with the pfMultipipe() call
to configure pfPipes that are not associated with a hyperpipe group. The num argument
to the pfMultipipe() function defines the total number of pfPipes to configure (including
those in hyperpipe groups).
Example 14-1, diagrammed in Figure 14-2, shows the configuration of a system with
three hyperpipe groups. The first hyperpipe group consists of three graphic pipes. The
506
007-1680-100
Using OpenGL Performer with a DPLEX
remaining two hyperpipe groups have two graphic pipes each. This example also
configures one non-hyperpipe group graphic pipe.
Example 14-1
Configuring a System with Three Hyperpipe Groups
pfInit();
pfMultipipe(8);
pfHyperpipe(3);
pfHyperpipe(2);
pfHyperpipe(2);
pfConfig();
/*
/*
/*
/*
/*
need eight pfPipes 3-2-2-1 */
pfPipes 0, 1, 2 are the first group */
pfPipes 3, 4 are the second group */
pfPipes 5, 6 are the third group */
construct the pfPipes */
If the target configuration includes only hyperpipe groups, it is not necessary to invoke
pfMultipipe(). OpenGL Performer correctly determines the number of pfPipes from the
pfHyperpipe() calls.
Using the Default Hyperpipe Mapping to Graphic Pipes
The pfPipes constructed by pfConfig() are ordered into a linear array and are selected
with the pfGetPipe() function. The pfPipes that are part of a hyperpipe group always
appear in this array before any non-hyperpipe group pfPipes.
The pfHyperpipe() function groups pfPipes together starting, by default, with pfPipe
number 0. In the following example, there are four pfPipes; the first two are combined
into a hyperpipe group:
pfMultipipe(4);
pfHyperpipe(2);
pfConfig();
OpenGL Performer maps each pfPipe to a graphic pipe, which is associated with a
specific X display, as shown in Figure 14-3:
Hyperpipe
pfPipe
0
1
2
3
Graphic pipe
0
1
2
3
Figure 14-3
007-1680-100
Single pipes
Default Hyperpipe Mapping to Graphic Pipes
507
14: Using Scalable Graphics Hardware
Using Nondefault Hyperpipe Mappings to Graphics Pipes
Each graphics pipe is associated with only one X screen. By default, OpenGL Performer
assigns each pfPipe to the screen of the default X display that matches the pfPipe index
in the pfPipe array; in other words, pfPipe(0) in the hyperpipe is mapped to X screen 0.
In most configurations, this default mapping is not sufficient. The second phase,
therefore, involves associating the configured pfPipes with the graphic pipes. This is
achieved through the pfPipeScreen() or pfPipeWSConnectionName() function on the
pfPipes of the hyperpipe group.
Example 14-2 shows, given the configuration in Example 14-1, how to map the pfPipes
to the appropriate screens. In this example, all of the graphic pipes are managed under
the same X display, that is, a different screen on the same display.
Example 14-2
Mapping Hyperpipes to Graphic Pipes
/* assign the single pfPipe to screen 0 */
pfPipeScreen(pfGetPipe(7), 0);
/* assign the pfPipes of hyperpipe group 0 to screens 1,2,3 */
for (i=0; i < 3; i++)
pfPipeScreen(pfGetPipe(i), i+1);
/* assign the pfPipes of hyperpipe group 1 to screens 4,5 */
for (i=3; i<5; i++)
pfPipeScreen(pfGetPipe(i), i+1);
/* assign the pfPipes of hyperpipe group 2 to screens 6,7 */
for (i=5; i<7; i++)
pfPipeScreen(pfGetPipe(i), i+1);
The following is a more complex example that uses GLXHyperpipeNetworkSGIX
returned from glXQueryHyperpipeNetworkSGIX() to configure the pfPipes. This
example is much more complete and is referred to in the following sections.
Example 14-3
More Complete Example: Mapping Hyperpipes to Graphic Pipes
int hasHyperpipe;
GLXHyperpipeNetworkSGIX* hyperNet;
int numHyperNet;
int i;
Display* dsp;
int numNet;
508
007-1680-100
Using OpenGL Performer with a DPLEX
int pipeIdx;
pfChannel* masterChan;
/* initialize Performer */
pfInit();
/* does this configuration support hyperpipe */
pfQueryFeature(PFQFTR_HYPERPIPE, &hasHyperpipe);
if (!hasHyperpipe) {
pfNotify(PFNFY_FATAL, PFNFY_RESOURCE, "no hyperpipe support");
exit(1);
}
/* query the network */
dsp = pfGetCurWSConnection();
hyperNet = glXQueryHyperpipeNetworkSGIX(dsp, &numHyperNet);
if (numHyperNet == 0) {
pfNotify(PFNFY_FATAL, PFNFY_RESOURCE, "no hyperpipes");
exit(1);
}
/*
* determine the number of distinct hyperpipe networks. network
* ids are monotonically increasing from zero. a value < 0
* is used to indicate pipes that are not members of any hyperpipe.
*/
for (i=0, numNet=-1; i<numHyperNet; i++)
if (numNet < hyperNet[i].networkId)
numNet = hyperNet[i].networkId;
numNet += 1;
/*
* configure all of the hyperpipes in the net
*
* NOTE * while it is possible to be selective about which hyperpipe(s)
* to configure, that is left as an exercise.
*/
for (i=0; i<numNet; i++) {
int count = 0;
int j;
for (j=0; j<numHyperNet; j++)
if (hyperNet[i].networkId == i) count++;
pfHyperpipe(count);
}
007-1680-100
509
14: Using Scalable Graphics Hardware
pfConfig();
/* associate pfPipes with screens */
for (i=0, pipeIdx=0; i<numNet; i++) {
int j;
for (j=0; j<numHyperNet; j++)
if (hyperNet[i].networkId == i)
pfPipeWSConnectionName(pfGetPipe(pipeIdx++),
hyperNet[i].pipeName);
}
/* construct the pfPipeWindows for each hyperpipe */
masterChan = NULL;
for (i=0, pipeIdx=0; i<numNet; i++) {
pfPipe* pipe;
pfPipeWindow* pwin;
pfChannel* chan;
PFVEC3 xyz, hpr;
pipe = pfGetPipe(pipeIdx);
pwin = pfNewPWin(pipe);
pfPWinName(pwin, "Hyperpipe Window");
/*
* void
* openPipeWindow(pfPipeWindow* pwin)
* {
*
pfPWinOpen(pwin);
* }
*/
pfPWinConfigFunc(pwin, openPipeWindow);
pfPWinFullScreen(pwin);
pfPWinMode(pwin, PFWIN_NOBORDER, 1);
pfPWinConfig(pwin);
chan = pfNewChan(pipe);
pfPWinAddChan(pwin, chan);
/*
* layout channels left to right in hyperpipe order. this
* ordering is arbitrary and should be redefined for the
* specific application.
*/
pfChanShare(chan,
510
007-1680-100
Using OpenGL Performer with a DPLEX
pfGetChanShare() | PFCHAN_VIEWPORT |
PFCHAN_SWAPBUFFERS | PFCHAN_SWAPBUFFERS_HW);
pfMakeSimpleChan(chan, 45);
pfChanAutoAspect(chan, PFFRUST_CALC_VERT);
xyz[0] = xyz[1] = xyz[2] = 0;
hpr[0] = (((numNet-1)*.5f)-i)*45.f;
hpr[1] = hpr[2] = 0;
pfChanViewOffsets(chan, xyz, hpr);
pfChanNearFar(.000001, 100000);
/*
* void
* drawFunc(pfChannel* chan, void* notUsed)
* {
*
pfClearChan(chan);
*
pfDraw();
* }
*/
pfChanTravFunc(PFTRAV_DRAW, drawFunc);
if (i == 0)
masterChan = chan;
else
pfAttachChan(masterChan, chan);
/* bump to the first pipe of the next hyperpipe */
pipeIdx += pfGetHyperpipe(pipe);
}
/*
* the next step is to construct the scene, attach it to
* masterChan and start the main loop. this bit of code
* is not included here since it follows other demonstration
* applications included elsewhere in the Programmer’s Guide.
*/
Configuring pfPipeWindows and pfChannels
The pfPipes grouped into a pfHyperpipe are indexed; the first pfPipe is pfPipe(0) and it
is referred to as the master pfPipe. Most actions taken on the hyperpipe group are
effected through this pfPipe; for example, all objects, such as pfPipeWindows and
pfChannels, are attached to the master pfPipe. OpenGL Performer automatically clones
007-1680-100
511
14: Using Scalable Graphics Hardware
all objects, except pfChannels, across all of the pfPipes in the pfHyperpipe, as shown in
Figure 14-4.
h
pfC
an
ne
l
p
p
fHy
erp
pfP
Master pipe
pfP
ipe
ipe
ipe
Cloned from
master pipe
pfPipeWindow
pfPipeWindow
Figure 14-4
Attaching Objects to the Master pfPipe
When constructing pfPipeWindows or pfChannels, the pfPipe argument should be the
master pfPipe. OpenGL Performer ensures that the constructed objects are cloned
(pfPipeWindows) or attached (pfChannels) as needed to the other pfPipes in the
hyperpipe group.
512
007-1680-100
Using OpenGL Performer with a DPLEX
With the exception of certain attributes, detailed in Table 14-1, OpenGL Performer
propagates attribute updates to the cloned pfPipeWindows when they occur. The
following is a list of pfPipeWindow functions for which the attributes do not propagate.
Table 14-1
007-1680-100
pfPipeWindow Functions That Do Not Propagate
C Function
C++ Member Function
pfPWinSwapBarrier()
setSwapBarrier()
pfPWinWSConnectionName()
setWSConnectionName()
pfPWinOverlayWin()
setOverlayWin()
pfPWinStatsWin()
setStatsWin()
pfPWinScreen()
setScreen()
pfPWinWSWindow()
setWSWindow()
pfPWinWSDrawable()
setWSDrawable()
pfPWinFBConfigData()
setFBConfigData()
pfPWinFBConfigAttrs()
setFBConfigAttrs()
pfPWinFBConfig()
setFBConfig()
pfPWinFBConfigId()
setFBConfigId()
pfPWinGLCxt()
setGLCxt()
pfPWinList()
setWinList()
pfPWinPVChan()
setPVChan()
pfPWinAddPVChan()
addPVChan()
pfPWinRemovePVChan()
removePVChan()
pfPWinRemovePVChanIndex()
removePVChanIndex()
pfBindPWinPVChans()
bindPVChans()
pfUnbindPWinPVChans()
unbindPVChans()
pfSelectPWin()
select()
pfAttachPWinWin()
attachWin()
513
14: Using Scalable Graphics Hardware
Table 14-1
pfPipeWindow Functions That Do Not Propagate (continued)
C Function
C++ Member Function
pfDetachPWinWin()
detachWin()
pfAttachPWin()
attach()
pfAttachPWinSwapGroup()
attachSwapGroup()
pfAttachPWinWinSwapGroup()
attachWinSwapGroup()
pfDetachPWinSwapGroup()
detachSwapGroup()
pfChoosePWinFBConfig()
chooseFBConfig()
When using any of the preceding interfaces within an application, set the appropriate
attribute in the cloned pfPipeWindow.
Clones
Clones are identified by an index value. The index of a clone matches that of the master
pfPipeWindow. This index is used to retrieve the clone pfPipeWindow from the other
pfPipes in the hyperpipe group. Example 14-4 sets the FBConfigAttrs for each of the
pfPipeWindows in the first hyperpipe group.
Example 14-4
Set FBConfigAttrs for Each pfPipeWindow
static int attr[] = {
GLX_RGBA,
GLX_DOUBLEBUFFER,
GLX_LEVEL, 0,
GLX_RED_SIZE, 8,
GLX_GREEN_SIZE, 8,
GLX_BLUE_SIZE, 8,
GLX_ALPHA_SIZE, 8,
GLX_DEPTH_SIZE, 16,
GLX_STENCIL_SIZE, 0,
GLX_ACCUM_RED_SIZE, 0,
GLX_SAMPLE_BUFFERS_SGIS, 1,
GLX_SAMPLES_SGIS, 4,
None
};
int numHyper = pfGetHyperpipe(pfGetPipe(0));
514
007-1680-100
Using OpenGL Performer with a DPLEX
for (i=0; i<numHyper; i++) {
/* get the first pfPipeWindow on pfPipe */
pfPipeWindow* pwin = pfGetPipePWin(pfGetPipe(i), 0);
pfPipeFBConfigAttrs(pwin, attr);
}
The current API has no support for directly querying the pfPipeWindow index within the
pfPipe. The only mechanism to determine an index value is to track it in the application
or search the pfPipeWindow list of the pfPipe. Example 14-5 performs such a search.
Example 14-5
Search the pfPipeWindow List of the pfPipe
/* search the master pfPipe pipe for the pfPipeWindow in pwin */
int pwinIdx;
int numPWins = pfGetPipeNumPWins(pipe);
for (i=0; i<numPWins; i++)
if (pfGetPipePWin(pipe) == pwin) break;
if (i == numPWins)
pfNotify(PFNFY_FATAL, PFNFY_PRINT, "oops!");
pwinIdx = i;
Synchronization
When working with pfPipeWindows, it is possible for some updates to occur within the
DRAW process. For this release (and possibly future releases) of OpenGL Performer,
these updates are not automatically propagated to the clone pfPipeWindows. It is the
responsibility of the application to ensure that the appropriate attributes are propagated
or that similar actions occur on the clones.
The CULL and DRAW stages of different pfPipes within a hyperpipe group can run in
parallel. For this reason, applications that assume a fixed pfChannel to pfPipe
relationship or maintain global configuration data associated with a pfChannel that is
updated in either the CULL or DRAW stages may fail. It is currently impossible (or at
least very difficult) to transmit information from the CULL or DRAW stages of one
pfPipe to another CULL or DRAW stage of another pfPipe within a hyperpipe group. All
changes should be affected by the APP stage.
Programming with Hyperpipes
Programming with hyperpipes, as described in the preceding sections, generally
involves the following steps:
007-1680-100
515
14: Using Scalable Graphics Hardware
1.
Configure the hyperpipe either on the fly or using a configuration file.
2. Map screens to hyperpipes, if necessary.
3. Allocate pfPipeWindow and pfChannels:
•
Create one pfPipeWindow for each pfHyperpipe.
•
Attach a pfPipeWindow to the master pfPipe.
•
Create a pfChannel for each pfHyperpipe.
4. Start the main loop (pfFrame()...pfSync()).
There are two additional requirements for DPLEX:
•
You cannot use single buffer visuals.
The DPLEX option uses the glXSwapBuffers() call as an indication to switch the
multiplexer. This logic is bypassed for single buffered visuals.
•
glXSwapBuffers() and pfSwapWinBuffers() functions must not be invoked outside
of the internal draw synchronization logic.
Because the pfuDownloadTexList() function with the style parameter set to
PFUTEX_SHOW calls glXSwapBuffers(), this feature must be disabled. (Simply set
the style parameter to PFUTEX_APPLY).
Also, the Perfly application displays a message at startup which also swaps the
buffers. Again, this function must be disabled when using hyperpipe groups. The
version of Perfly that ships with performer_demo correctly disables these
features.
Each pfPipe software rendering pipeline runs at a fraction of the target frame rate as
defined by pfFrameRate(). The fraction is 1/(number of pipes in hyperpipe group). For
example, if there are two pfPipes in the pfHyperpipe, each pfPipe runs at one half of the
pfFrameRate(). Although the CULL and DRAW stages run at a slower rate, the APP
stage must run at the target frame rate.
516
007-1680-100
Using OpenGL Performer with an SGI Scalable Graphics Compositor
Using OpenGL Performer with an SGI Scalable Graphics Compositor
This section gives a brief overview of the SGI Scalable Graphics Compositor and how to
use it with OpenGL Performer. For more information on the compositor, including the
details of the hardware setup, see the document SGI InfinitePerformance: Scalable Graphics
Compositor User’s Guide.
Note: The compositor is currently supported on InfinitePerformance, Onyx4, and Prism
graphics systems.
This section contains the following subsections:
•
“How the Compositor Functions” on page 517
•
“The pfCompositor Class” on page 519
•
“Querying the System for Hardware Compositors” on page 519
•
“Creating a pfCompositor” on page 520
•
“Querying pfCompositors” on page 522
•
“Load Balancing” on page 524
•
“Setting Compositor Modes” on page 525
•
“Querying Compositor Modes” on page 528
•
“Managing Screen Space, Channel Clipping, and Antialiasing” on page 529
How the Compositor Functions
The compositor receives two to four input signals and outputs a single signal either in
analog or digital format. Hence, it can handle spatial composition of four inputs which
enables multiple pipes to contribute to a single output. Four different composition
schemes are available:
007-1680-100
•
Vertical stripes
•
Horizontal stripes
•
2D tiles
•
Cut-ins
517
14: Using Scalable Graphics Hardware
Figure 14-5 illustrates the various hardware composition schemes.
Vertical stripes
Horizontal stripes
2D tiles
Cut-ins
Figure 14-5
Hardware Composition Schemes
The following items are noteworthy regarding the compositor’s capabilities:
•
In addition to the spatial composition modes shown in Figure 14-5, the compositor
provides applications with the means to do full-scene antialiasing (FSAA) in
hardware. This capability stems from the following feature: for every output pixel,
the compositor averages all values from all the pipes.
•
Stereo support is not provided explicitly through the compositor, but
OpenGL Performer does allow you to structure your application to do so.
Note: For more information on the current limitations and anomalies associated with the
use of the SGI Scalable Graphics Compositor, refer to the hardware documentation.
518
007-1680-100
Using OpenGL Performer with an SGI Scalable Graphics Compositor
The pfCompositor Class
A compositor is a hardware device that takes a number of video inputs and combines
them to produce a single video output. The video inputs can be divided spatially or
blended together to form one output. OpenGL Performer uses the pfCompositor class to
support compositors. The pfCompositor class transparently distributes rendering across
multiple hardware pipes and combines their outputs by either feeding them to a
hardware compositor device or through software composition.
Several different spatial composition modes are supported, as well as an antialias mode
in which channel frustums on composited pipes are slightly jittered and the outputs
blended together by the hardware compositor.
The pfCompositor class also supports dynamic load balancing. When enabled, the
spatial subdivision of the compositor inputs will be updated on each frame based on the
load of each contributing pfPipe. Load balancing is disabled by default but can be
enabled through setMode().
Querying the System for Hardware Compositors
During initialization, the pfCompositor class will perform a system topology query to
determine the availability of hardware compositors. The results of this query can be
examined by the application through the static methods listed in Table 14-2.
Table 14-2
007-1680-100
Methods for Querying the System for Hardware Compositors
Methods
Description
getNumHWCompositors()
Returns the number of available hardware compositors
found on the system.
getHWCompositorNetworkId()
Returns the network ID of the cth hardware
compositor, or –1 if c is not a valid index.
getHWCompositorNumInputs()
Returns the number of inputs physically connected to
the cth hardware compositor. Each input can either be
a single pipe or the output of another hardware
compositor. If c is not a valid index, –1 is returned.
519
14: Using Scalable Graphics Hardware
Table 14-2
Methods for Querying the System for Hardware Compositors (continued)
Methods
Description
getHWCompositorInputType()
Returns PFCOMP_INPUTTYPE_PIPE if the ith input of
the cth hardware compositor is a single pipe, or
PFCOMP_INPUTTYPE_COMPOSITOR if it is another
hardware compositor. If c is not a valid index, 0 is
returned.
getHWCompositorInputPipeName()
Returns the string identifying the display of that pipe
(for example, ":0.0") if the ith input of the cth
hardware compositor is a single pipe. If c or i is not a
valid index or if the ith input of the cth compositor is
not a single pipe, NULL is returned.
getHWCompositorInputNetworkId()
Returns its network id if the ith input of the cth
hardware compositor is a compositor. If c or i is not a
valid index or if the ith input of the cth compositor is
not a compositor, –1 is returned.
getHWCompositorPfCompositor()
Returns a pointer to the pfCompositor object managing
the cth hardware compositor if one exists and c is a
valid index. Otherwise, NULL is returned.
Creating a pfCompositor
A pfCompositor is created through new pfCompositor(netId), where netId is the network
ID for the hardware compositor device that will be managed by the new pfCompositor.
If netId is PFCOMP_SOFTWARE, no hardware compositor device will be involved, and
composition will be carried out through software readbacks.
This chapter refers to pfCompositors utilizing software composition as software
compositors while pfCompositors associated with a hardware compositor device will be
referred to as hardware compositors.
520
007-1680-100
Using OpenGL Performer with an SGI Scalable Graphics Compositor
You can use the methods described in Table 14-3 to configure pfCompositors that have
been created.
Table 14-3
Methods Used in Creating pfCompositors
Methods
Description
getNetworkId()
Returns the network id identifying the hardware compositor device
managed by a pfCompositor object or PFCOMP_SOFTWARE if the
pfCompositor uses software composition.
addChild(pipe_name)
Adds a pipe child (input) to a pfCompositor. If the pfCompositor is
a hardware compositor, pipe_name must match the display string
of one of the hardware pipes physically connected to the compositor
device. If the compositor is a software compositor, then pipe_name
can be any valid display string. The pfPipes configured by the
software compositor will be created on the specified displays
through calls to pfPipe::setWSConnectionName(). The
pipe_name value can also be "" (empty string) for software
compositors; in this case, pipes will be created on the default screens
(:0.0, :0.1, and so on).
addChild(comp)
Adds a compositor child to a pfCompositor, creating a compositor
hierarchy. Currently only hardware compositor parents and
software compositor children are supported. Care must be taken in
configuring compositor hierarchies to ensure that the first child of
the software compositor child (its master pipe) is physically
connected to the compositor device managed by the parent
pfCompositor.
autoSetup()
Configures a pfCompositor with the desired number of inputs. For
hardware compositors, num_inputs cannot exceed the number of
inputs physically connected to the hardware device. A zero or
negative value for num_inputs will cause all physically connected
inputs to be configured.
For software compositors, num_inputs will be clamped to the
number of available hardware pipes. If num_inputs is less than one,
all available hardware pipes on the system will be configured.
Note that autoSetup() will take no action at all if pfCompositor
already has one or more children. The call autoSetup(–1) is made
within a pfConfig for all pfCompositor objects in order to
automatically configure them if the application has not done so
already.
007-1680-100
521
14: Using Scalable Graphics Hardware
Querying pfCompositors
The methods described in Table 14-4 can be used to query compositors.
Table 14-4
522
Methods for Querying pfCompositors
Method
Description
getNumChildren()
Returns the number of children (inputs) that have been added to
a pfCompositor. Each child can either be a single pipe or a
pfCompositor.
getChildType()
Returns PFCOMP_INPUTTYPE_PIPE if the ith child is a single
pipe or PFCOMP_INPUTTYPE_COMPOSITOR if it is a
pfCompositor. If i is not a valid index, 0 is returned.
getChildCompositor()
Returns a pointer to the ith child of a pfCompositor if the child is
a pfCompositor. If i is not a valid index, or if the ith child is not a
compositor, NULL is returned.
getChildPipe()
Returns a pointer to the ith child of a pfCompositor if the child is
a single pipe. This can only be called after pfConfig().
getChildPipeName()
Returns the display string for the ith child of a pfCompositor if the
child is a single pipe. Note that for software compositors,
getChildPipeName() returns an empty string ("") for all children
unless a display string was explicitly assigned by the application
through a call to addChild(pipe_name).
getChildPipeId()
Returns the OpenGL Performer ID of the pipe child and should
only be called after pfConfig().
getParent()
Returns a pointer to a pfCompositor parent (another
pfCompositor) if the first has been added to the latter as a child
through a call to addChild(comp). If a pfCompositor has no
parent, NULL is returned.
setNumActiveChildren()
Sets the number of active children. Not all configured children
must contribute to the composited image at all times. There must
be at least one active child and no more than the total number of
configured children. Note that children are activated from first to
last; this means that when there are n active children, these will be
children 0 to (n-1); thus, child 0 is always active.
getNumActiveChildren()
Returns the number of currently active children.
007-1680-100
Using OpenGL Performer with an SGI Scalable Graphics Compositor
Table 14-4
007-1680-100
Methods for Querying pfCompositors (continued)
Method
Description
getMasterPipe()
Returns a pointer to the master pfPipe for a pfCompositor. The
master pipe is the pipe that the application should use to create a
pipe window and one or more channels. The pfPipeWindows and
pfChannels are created automatically on all other composited
pipes (slave pipes) by the pfCompositor class. In a single-tier
compositor (one with no compositor parent and no compositor
children), the master pipe will be its first child. In a compositor
hierarchy, all pfCompositors will share a single master pipe.
getMasterPipeId()
Returns the OpenGL Performer pipe ID of pfCompositor's master
pipe. Each pfCompositor object maintains a list of all the pfPipes
contributing to its output. This includes all single-pipe children, as
well as all single pipes connected to compositor children.
getNumPipes()
Returns the total number of pfPipes contributing to a
pfCompositor. For a single-tier compositor, this value is equal to
the number of its children. In a compositor hierarchy, pipes
contributing to leaf compositors (bottom of the hierarchy) also
contribute to the root compositor; therefore, if called on the root
compositor of a compositor hierarchy, getnumPipes() returns the
total number of pipes in hierarchy.
getPipe()
Returns a pointer to the pth pfPipe in a pfCompositor's pipe list.
If p is an invalid index, NULL is returned.
getPWin()
Returns a pointer to the pfPipeWindow on the pth pipe in a
pfCompositor's pipe list. Currently only one (full-screen) pipe
window is supported on composited pipes. If p is an invalid index,
NULL is returned.
getChan()
Returns a pointer to the cth pfChannel on the pth pipe in a
pfCompositor's pipe list. If p or c are invalid indexes, NULL is
returned.
getRoot()
Returns a pointer to the pfCompositor at the root of the
compositor hierarchy to which compositor belongs. For a
parent-less pfCompositor, getRoot() returns a pointer to the
compositor itself. For a compositor child, getRoot() returns
parent->getRoot.
523
14: Using Scalable Graphics Hardware
In addition to the methods shown in Table 14-4, the static methods shown in Table 14-5
are also available.
Table 14-5
Static Methods for Querying pfCompositors
Method
Description
getNumCompositors()
Returns the number of pfCompositor objects in this list.
getCompositor()
Returns a pointer to the ith pfCompositor object from the global
list of pfCompositors. The pfCompositors are added to this list in
the order of creation.
getNumCompositedPipes()
Returns the total number of pfPipes that are (or will be) managed
by pfCompositor objects. This is known for certain only after a
pfConfig() call because, until then, pipes may be added to
existing compositors. However, if called before pfConfig(), this
method will attempt to make a reasonable guess by assuming that
pfCompositors with no explicitly assigned children will end up
being (automatically) configured with all the inputs that are
physically connected to them.
This can be useful when an application creates one or more single
pipes in addition to pipes managed by pfCompositors. In such
cases, the application is required to make a call to pfMultipipe()
to provide the total number of pipes to be created. Method
getNumCompositedPipes() returns the total number of
composited pipes to which the desired number of single pipes
may be added.
Load Balancing
A pfCompositor requires a pfLoadBalance object for carrying out load balancing
computations. The pfLoadBalance class determines the resulting workload for each of
the compositor's children. The behavior can be customized by subclassing
pfLoadBalance and overriding the appropriate methods. A pfCompositor can use a
customized pfLoadBalance object specified through the setLoadBalancer() method. If a
load balancer is not specified, one will be automatically created and used. The method
getLoadBalancer() returns a pointer to the pfLoadBalancer object used by the
compositor.
524
007-1680-100
Using OpenGL Performer with an SGI Scalable Graphics Compositor
The two methods in Table 14-6 can be used to control the transition between load
balancing states.
Table 14-6
Methods to Control the Load Balancing Transitions
Method
Description
setVal()
Accepts only PFLOAD_COEFF and passes the value to the pfLoadBalance class. This
coefficient determines how quickly the balancer transitions from the current state to
the desired balanced state. This load balancing filter coefficient should be in the
range [0..1]. The smaller its value, the slower load balancing follows pipe loading,
and the less noise-sensitive it is.
getVal()
Accepts only PFLOAD_COEFF and returns the current value of the filter coefficient
used by the pfLoadBalance object associated with the pfCompositor.
For more information, see the pfLoadBalance man page.
Setting Compositor Modes
The method setMode() accepts the following tokens as its first argument:
PFLOAD_BALANCE
Enables or disables dynamic load balancing. The
second argument must be PF_ON or PF_OFF.
PFCOMP_CLIPPING
Enables or disables channel clipping for all channels
on all pipes managed by this compositor. By default,
channel clipping is enabled, and the viewports of
pfChannels in composited pipes are clipped to the
screen region assigned to each pipe by the
compositor. If channel clipping is disabled, all pipes
will render all channels in their full (original) size.
Note that clipping is not carried out when in antialias
mode.
PFCOMP_SWAPBARRIER
007-1680-100
Specifies the swap barrier to which pipes
contributing to pfCompositor should bind. The
second argument should be a valid swap barrier ID
(see the glXQueryMaxSwapBarriersSGIX man
page). By default, all pfCompositors will bind to
swap barrier 1 if the swap barrier extension is
525
14: Using Scalable Graphics Hardware
supported. Binding to a swap barrier can be disabled
by passing a value smaller than 1. If the specified
barrier_id is out of range, the call to setMode()
has no effect.
PFCOMP_COMPOSITION_MODE
Specifies the composition mode used by the
pfCompositor. The second argument can be one of the
following:
PFCOMP_COMPMODE_HORIZ_STRIPES
PFCOMP_COMPMODE_VERT_STRIPES
PFCOMP_COMPMODE_LEFT_TILES
PFCOMP_COMPMODE_RIGHT_TILES
PFCOMP_COMPMODE_BOTT_TILES
PFCOMP_COMPMODE_TOP_TILES
PFCOMP_COMPMODE_ANTIALIAS
All composition modes are valid for any number of active children.
Figure 14-6 illustrates how one to four inputs are laid out for
PFCOMP_COMPMODE_HORIZ_STRIPES mode.
PFCOMP_COMPMODE_HORIZ_STRIPES
1
0
1
0
Figure 14-6
2
0
3
2
1
0
Horizontal Stripes (pfCompositor Mode)
Figure 14-7 illustrates how one to four inputs are laid out for
PFCOMP_COMPMODE_VERT_STRIPES mode.
526
007-1680-100
Using OpenGL Performer with an SGI Scalable Graphics Compositor
PFCOMP_COMPMODE_VERT_STRIPES
0
Figure 14-7
0
1
0
1
2
0
1
2
3
Vertical Stripes (pfCompositor Mode)
Figure 14-8 illustrates how one to four inputs are laid out for
PFCOMP_COMPMODE_LEFT_TILES mode.
PFCOMP_COMPMODE_LEFT_TILES
0
Figure 14-8
0
1
2
1
1
0
3
0
2
Left Tiles (pfCompositor Mode)
Figure 14-9 illustrates how one to four inputs are laid out for
PFCOMP_COMPMODE_RIGHT_TILES mode.
PFCOMP_COMPMODE_RIGHT_TILES
2
0
0
1
3
0
1
Figure 14-9
1
0
2
Right Tiles (pfCompositor Mode)
Figure 14-10 illustrates how one to four inputs are laid out for
PFCOMP_COMPMODE_BOTT_TILES mode.
007-1680-100
527
14: Using Scalable Graphics Hardware
PFCOMP_COMPMODE_BOTT_TILES
1
1
2
2
3
0
0
Figure 14-10
0
0
1
Bottom Tiles (pfCompositor Mode)
Figure 14-11 illustrates how one to four inputs are laid out for
PFCOMP_COMPMODE_TOP_TILES mode.
PFCOMP_COMPMODE_TOP_TILES
0
1
3
2
0
0
Figure 14-11
1
2
0
1
Top Tiles (pfCompositor Mode)
Querying Compositor Modes
The method getMode() can be called to query the compositor mode. You can use the
following tokens:
528
PFLOAD_BALANCE
The returned value is 1 if dynamic load balancing is
enabled or 0 if it is disabled.
PFCOMP_CLIPPING
The returned value is 1 if channel clipping is enabled
or 0 if it is disabled.
007-1680-100
Using OpenGL Performer with an SGI Scalable Graphics Compositor
PFCOMP_SOFTWARE
The returned value is 1 if pfCompositor uses software
composition or 0 if pfCompositor controls a
hardware compositor device.
PFCOMP_SWAPBARRIER
The returned value is the index of the swap barrier to
which pipes will bind (or have bound). If binding to
swap barriers has been (or will be) skipped, return
value is 0.
PFCOMP_COMPOSITION_MODE
The returned value is the current composition mode
used by the pfCompositor and can be one of the
following:
PFCOMP_COMPMODE_HORIZ_STRIPES
PFCOMP_COMPMODE_VERT_STRIPES
PFCOMP_COMPMODE_LEFT_TILES
PFCOMP_COMPMODE_RIGHT_TILES
PFCOMP_COMPMODE_BOTT_TILES
PFCOMP_COMPMODE_TOP_TILES
PFCOMP_COMPMODE_ANTIALIAS
Managing Screen Space, Channel Clipping, and Antialiasing
You can use the methods described in Table 14-7 to manage screen space, channel
clipping, and antialiasing.
Table 14-7
007-1680-100
Methods for Managing Screen Space, Channel Clipping, and Antialiasing
Method
Description
setViewport()
Specifies the screen-space bounds of the region managed by a
pfCompositor. The viewports assigned to all pipes managed by this
compositor will be clipped to this region. The default viewport for a
pfCompositor is 0.0, 1.0, 0.0, 1.0 (the whole screen). Do not call
setViewport() for compositor children in compositor hierarchies; use
setChildViewport() on the parent compositor instead.
getViewport()
Returns the screen-space bounds of the region managed by a
pfCompositor.
529
14: Using Scalable Graphics Hardware
Table 14-7
Methods for Managing Screen Space, Channel Clipping, and Antialiasing
Method
Description
setChildViewport()
Specifies the screen-space bounds of the 2D region assigned to the ith
child of the pfCompositor. The specified viewport is automatically
clipped to the viewport of the compositor and aligned horizontally to a
four-pixel boundary (required by hardware compositor devices). The
default viewports of a pfCompositor's children are determined based on
the number of active children and on the current composition mode; see
the preceding description for setMode(). Note that when dynamic load
balancing is active, setting children viewports through
setChildViewport() will have no affect.
getChildViewport()
Returns the screen-space bounds of the 2D region managed by the ith
child of the pfCompositor. This region will always be contained by the
viewport of the pfCompositor itself.
setChannelClipped()
Specifies whether channel clipping should be enabled for the ith
channel. Channel clipping is enabled on all channels by default.
Channel clipping is not performed if it is globally disabled through a
call to setMode(). Disabling clipping on a pfChannel can be useful in
certain situations; for example, the GUI channel for Perfly has clipping
disabled and is rendered entirely on the master pipe.
getChannelClipped()
Returns 1 if channel clipping is enabled for the ith channel if i is a valid
index or 0, otherwise.
setAntialiasJitter()
Specifies the jitter pattern to be used for antialias composition when
there are n active children. The parameter jitter must point to an array
of floats containing 2*n values, specifying subpixel offsets (horizontal
and vertical) for each of the n contributing inputs. A pfCompositor
maintains a list of jitter patterns to be used for antialias mode,
depending on the number of active children. A jitter pattern is encoded
as an array of subpixel offsets with two floats (horizontal and vertical
offset) for each contributing child.
getAntialiasJitter()
Returns the jitter pattern to be used for antialias composition when
there are n active children. The parameter jitter must point to an array
of floats (with at least 2*n elements), which contains the queried jitter
values.
Note: All viewports are specified in normalized screen coordinates with 0.0,0.0 as the
bottom-left corner of the screen and 1.0,1.0 as the top-right corner.
530
007-1680-100
Using OpenGL Performer with GPUs
Using OpenGL Performer with GPUs
GPUs are used widely in commodity graphics hardware and also in graphics platforms
like Onyx4 or Prism systems. OpenGL Performer supports GPU programming through
the use vertex programs and fragment programs.
Vertex programs are used by the GPU to modify various parameters of each vertex.
Similarly, fragment programs are used to modify the color and depth value of each
fragment (pixel) as it is being rendered. A description of vertex and fragment program
instruction sets is beyond the scope of this guide. You can find a description of these
instruction sets in the OpenGL extension registry at
http://oss.sgi.com/projects/ogl-sample/registry/ under GL_ARB_vertex_program
and GL_ARB_fragment_program.
This chapter describes how you can use GPU programs in OpenGL Performer in the
following sections:
•
“The pfGProgram Class” on page 532
•
“The pfGProgramParms Class” on page 534
•
“The pfVertexProgram and pfFragmentProgram Classes” on page 535
The pfGProgram Class
OpenGL Performer implements GPU programming through the general class
pfGProgram. This class allows you to set GPU programs, vertex programs and fragment
programs.
The function pfNewGProgram() creates and returns a handle to a pfGProgram. The
parameter arena specifies a malloc() arena out of which the pfGProgram is allocated or
the value NULL specifies allocation off the process heap. pfGPrograms can be deleted
with pfDelete().
The call new(arena) allocates a pfGProgram from the specified memory arena or from
the heap if arena is NULL. The new() call allocates a pfGProgram from the default
memory arena (see function pfGetSharedArena() in the pfSharedMem(3pf) man page).
Like other pfObjects, pfGPrograms cannot be automatically created statically on the
stack or in arrays. pfGPrograms should be deleted with pfDelete() rather than the delete
operator.
007-1680-100
531
14: Using Scalable Graphics Hardware
The function pfGetGProgramClassType() returns the pfType* for the class
pfGProgram. The pfType* returned by pfGetGProgramClassType() is the same as the
pfType* returned by invoking pfGetType(), the virtual function getType() on any
instance of class pfGProgram. Because OpenGL Performer allows subclassing of built-in
types, when decisions are made based on the type of an object, it is usually better to use
pfIsOfType(), the member function isOfType(), to test if an object is of a type derived
from an OpenGL Performer type rather than to test for strict equality of the types.
A pfGProgram is a sequence of assembly-like instructions. You can specify the
instructions in two ways:
•
In a string with new line characters separating instructions
•
In a text file
If the program is specified in a string, you use the function pfGProgramProgram() or
pfGProgramProgramLen(). The first parameter of each is the string defining the
program. In the second function, you can specify the length when you want to load only
part of the string.
If the program is loaded from a text file, you use the function
pfGProgramLoadProgram().
Using the function pfGProgramApplypfGProgram(), you can apply the pfGProgram
but only in the draw process. Once the pfGProgram has been applied, you can query its
state using the following functions:
pfGetGProgramProgramLength()
Returns the number of instructions of the
program.
pfGetGProgramNativeProgramLength() Returns the number of instructions used by
the specific GPU.
pfGProgramIsValidpfGProgram()
Returns 1 if the program has been
successfully loaded into the GPU.
You should not use a pfGProgram directly but one of its subclasses. There are two classes
of specific GPU programs subclassed from pfGProgram: pfVertexProgram and
pfFragmentProgram. A pfVertexProgram or a pfFragmentProgram is set in a pfGeoState
and is enabled in a pfGeoState. The user parameters for the vertex and fragment
programs can be defined using the class pfGProgramParams, which is described in the
following section.
532
007-1680-100
Using OpenGL Performer with GPUs
For sample code, see the following file:
/usr/share/Performer/src/pguide/libpf/C++/gprogram.C
(IRIX and Linux)
%PFROOT\Src\pguide\libpf\C++\gprogram.cxx
(Microsoft Windows)
The pfGProgramParms Class
The pfGProgramParms is a class that is used to store parameters of GPU programs,
specifically of pfVertexPrograms and pfFragmentPrograms. The function
pfNewGProgramParms() creates and returns a handle to a pfGProgramParms. The
parameter arena specifies a malloc() arena out of which the pfGProgram is allocated or
NULL for allocation off the process heap. You can delete pfGPrograms with pfDelete().
A pfGProgramParms is a set of indexed quadruples of floating point values that are used
as parameters for vertex and fragment programs. You can specify the values using the
function pfGPParamsParameters() which has the following syntax:
pfGPParamsParameters(pfGProgramParms* gpparams, int index, int type, int count, void* ptr);
The parameter index specifies the first index of the specified parameters (the index by
which the parameters are accessed in the GPU program) and the the parameter count
specifies how many indices will be set.
The parameter type may be one of the following:
PFGP_FRAGMENT_LOCAL
Local parameters of a single fragment program.
PFGP_FRAGMENT_ENV
Environment parameters. Shared between all
fragment programs.
PFGP_VERTEX_LOCAL
Environment parameters of a single vertex program.
PFGP_VERTEX_ENV
Environment parameters. Shared between all vertex
programs.
The pointer ptr points to the parameter data.
007-1680-100
533
14: Using Scalable Graphics Hardware
Using the following functions, you can query the existing parameters in a
pfGProgramParms:
pfGetGPParamsNumParameters()
Returns the number of parameters.
pfGetGPParamsParameters()
Returns the parameters in the order of their
specification.
pfGetGPParamsParametersByIndex()
Returns the parameters by the index by
which the parameters are accessed.
You can apply the pfGProgramParms using pfGProgramParamsApply() but only in the
draw process. If you modify the pfGProgramParms after they have been applied, you
must call pfGProgramParamsUpdate() for the change to take effect.
A pfGProgramParms is set in a pfGeoState. Each pfGeoState can have one
pfGProgramParms of each of the four types, two for the pfVertexProgram associated
with the pfGeoState and two for the pfFragmentProgram.
The pfVertexProgram and pfFragmentProgram Classes
The pfVertexProgram and pfFragmentProgram classes are derived from the class
pfGProgram. These subclasses do not add any new methods. A vertex program or a
fragment program is used by the GPU to modify various parameters of each vertex or
fragment (pixel), respectively. The GPU allows you to specify a sequence of
floating-point 4-component operations that are executed for each vertex or fragment.
These operations transform an input set of per-vertex or per-fragment parameters to
another set of per-vertex or per-fragment parameters.
A vertex program replaces the standard OpenGL set of lighting and texture coordinate
generation modes. Consequently, the vertex program must take care of the basic
transformation of vertex coordinates to the screen coordinates, the generation of texture
coordinates, and the application of the lighting equation. This programming model
allows you to modify the position of each vertex, producing, for example, a displacement
mapping.
Similar to a vertex program, a fragment program replaces the standard OpenGL set of
texture and fog application modes. The fragment program has to access the textures and
to modulate the resulting color according to the fog equation, if necessary. This
programming model allows you to modify the resulting color and depth of each pixel,
making it possible, for example, to apply a complex per-pixel shading.
534
007-1680-100
Using OpenGL Performer with GPUs
You can find the instruction sets for vertex and fragment programs in the OpenGL
extension registry at http://oss.sgi.com/projects/ogl-sample/registry/ under the
GL_ARB_vertex_program.
As subclasses of pfGProgram, pfVertexPrograms and pfFragmentPrograms can use the
management methods of a pfGProgram to set, load, and apply programs. Section “The
pfGProgram Class” on page 532 describe these methods.
You set and enable pfVertexPrograms and pfFragmentPrograms in a pfGeoState. As
described in section “The pfGProgramParms Class” on page 534, the user parameters for
GPU programs can be defined using the pfGProgramParms class.
For sample code, see the following file:
/usr/share/Performer/src/pguide/libpf/C++/gprogram.C
(IRIX and Linux)
%PFROOT\Src\pguide\libpf\C++\gprogram.cxx
(Microsoft Windows)
007-1680-100
535
14: Using Scalable Graphics Hardware
536
007-1680-100
Chapter 15
15. ClipTextures
As CPUs get faster and storage gets cheaper, applications are moving away from scenes
with small, synthetic textures to large textures, taken from real environments, giving the
viewer realistic renderings of actual locations.
There has customarily been a trade-off between the complexity of a texture and the area
it covers: if a texture covers a large area, its resolution must be limited so that it can fit
into texture memory; high-resolution textures are limited to small regions for the same
reason.
A cliptexture allows you to circumvent many of these system resource restrictions by
virtualizing MIPmapped textures. Only those parts of the texture needed to display the
textured geometry from a given location are stored in system and texture memory.
OpenGL Performer provides support for this technique, called cliptexturing, as a subclass
of a pfTexture called a pfClipTexture. This functionality allows you to display textures
too large to fit in texture memory or even in system memory; you can put the entire world
into a single texture.
OpenGL Performer supports texture load management from disk to system memory and
from system to texture memory, synchronizing clipped regions with the viewpoint, and
with the many other tasks needed to virtualize a texture relative to the viewer location.
This chapter describes cliptextures in the following parts:
007-1680-100
•
“Overview” on page 538
•
“Cliptexture API” on page 553
•
“Preprocessing ClipTextures” on page 553
•
“Cliptexture Configuration” on page 556
•
“Configuration API” on page 557
•
“Post-Scene Graph Load Configuration” on page 579
•
“Manipulating Cliptextures” on page 587
•
“Using Cliptextures” on page 601
537
15: ClipTextures
Overview
Cliptexturing avoids the size limitations of normal MIPmaps by clipping the size of each
level of a MIPmap texture to a fixed area, called the clip region. A MIPmap contains a
range of levels, each four times the size of the previous one. If the clip region is larger
than a particular level, the entire level is kept in texture memory. Levels larger than the
clip region are clipped to the clip region’s size. The clip region is set by the application,
trading off texture memory consumption against image quality. The clip region size is set
through the clip size, which is the length of the clip regions’s sides (in texels).
Clip size
Clip region
Entire level in
texture memory
Figure 15-1
Cliptexture Components
The clip region positioned so as to be centered about the clip center, or as close as possible
to the clipcenter while remaining entirely within the cliptexture. The clipcenter is set by
the application, usually to the location on the texture corresponding to the location
closest to the viewer on the cliptextured geometry. The clipcenter is specified in texel
coordinates, which is the texture coordinates (s and t values, ranging from 0.0 to 1.0,
scaled by the dimensions of the finest level of the cliptexture, level 0).
538
007-1680-100
Overview
Cliptexture Levels
Texture memory contains the MIPmap levels, the larger ones clipped to the clip region
size; the rectangle of texture memory corresponding to each clipped level is called a tex
region. As the viewer moves relative to the cliptextured geometry, the clipcenter must be
updated. When this happens, the clipped MIPmap levels must have their texture data
updated, in order to represent the area closest to the center. This updating usually must
happen every frame, and is done by OpenGL Performer image caches.
To facilitate loading only portions of the texture at a time, the texture data must first be
subdivided into a contiguous set of rectangular areas, called tiles. These tiles can then
loaded individually from disk into texture memory.
Texture memory must be loaded from system memory; it can’t be loaded directly from
disk. In order to improve the performance of texel downloading, the region in system
memory is made larger than the destination texture memory and organized into a
lookahead cache, called the mem region.
Clip region
Mem region
Tex region
Entire level in
texture memory
Figure 15-2
007-1680-100
Image Cache Components
539
15: ClipTextures
Image caches must know three things in order to update clipped texture levels:
•
Where and how the data is stored on disk, so they can retrieve it,
•
Location and size of system memory cache, called the mem region,
•
The texture memory they are responsible to update when the cilpcenter moves (the
tex region).
Cliptexture Assumptions
For the cliptexture algorithm to work seamlessly, applications must abide by the
following assumptions:
•
An application can only view a clip region’s worth of high resolution texel data on
its textured geometry from any viewpoint.
•
The application views the texture from one location at a time. Multiple views
require multiple cliptextures.
•
The viewer must move smoothly relative to the cliptextured geometry; no
“teleporting” (abrupt changes in position).
Given these assumptions, your application can maintain a high-resolution texture by
keeping only those parts of the texture closest to the viewer in texture memory; the
remainder of the texture is on disk and cached in system memory.
Why Do These Assumptions Work?
Only the textured geometry closest to the viewer needs a high-resolution texture. Far
away objects are smaller on the screen, so the texels used on that object also appear
smaller (cover a smaller screen area). In normal MIPmapping, coarser MIPmap levels are
chosen as the texel size gets smaller relative to the pixel size. These coarser levels contain
less texels, since each texel covers a larger area on the textured geometry.
Cliptextures take advantage of this fact by storing only part of each large MIPmap level
in texture memory, just enough so that when you look over the geometry, the MIPmap
algorithm starts choosing texels from a lower level (because the texels are getting small
on the screen) before you run out of texels on the clipped level. Because coarser levels
have texels that cover a larger area, at a great enough distance, MIPmapping is choosing
texels from the unclipped, smaller levels.
540
007-1680-100
Overview
When a clip size is chosen, cliptexture levels can be thought of as belonging to one of two
categories:
•
Clipped levels, which are texture levels that are larger than the clip size.
•
Non-clipped levels, which are small enough to fit entirely within the clip region.
The non-clipped levels are viewpoint independent; each non-clipped texture level is
complete. Clipped levels, however, must be updated as the viewer moves relative to the
textured geometry.
Image Cache
The image cache organizes its system memory as a grid of fixed size texture tiles. This
grid of texture data forms a lookahead cache, called the mem region. The cache
automatically anticipates texture download requirements, updating itself with texture
tiles it expects to use soon.
Image caches update texture memory by transferring image data from disk files. The
data is transferred in two steps. Data is moved from disk files a tile at a time into the mem
region in system memory. The mem region is updated so that it always contains the
image data corresponding to the tex region and its immediate surroundings. The border
of extra surrounding data allows the image cache to update the tex region as necessary
without having to wait for tiles to be loaded into the mem region from disk.
The image cache also contains a tex region, the rectangle of texel data in a given level’s
texture memory. This rectangle of data is in texture memory, and is being updated from
a corresponding rectangle of data in the memregion. As the center moves, the tex region
being loaded into texture memory can get close to the edge of the mem region. When this
happens, tiles in the mem region are updated with new data from disk so that the tex
region is moved closer to the center of the image data.
007-1680-100
541
Tex region
Mem region
Disk
System memory
Texture memory
15: ClipTextures
Disk files
Figure 15-3
Mem Region Update
As the center moves, the clipped region on each clipped level of the image cache shifts
position. The clipped regions on each level move at different rates; each coarser level only
moves at one half the speed of the level above it. The image cache reflects the change on
its level by tracking the position of the clipped region with its tex region. Data in texture
memory must be updated to match the texel data in the translated tex region.
This updating is done by copying rectangles of texel data from the shifted tex region area
in the mem region to the appropriate locations in texture memory. The amount of
updating is minimized by only updating the portions of the texture memory that actually
need new data. The majority of the tex region data only has to shift position in texture
memory; this is done by translating texture coordinates, and taking advantage of the
wrap mode when accessing texels from texture memory.
542
007-1680-100
Tex region
Tex region update
Mem region
Disk
System memory
Texture memory
Overview
Disk files
Figure 15-4
Tex Region Update
By loading textures to system memory before they are needed in texture memory, the
latency caused by waiting for tiles downloading from a disk is reduced.
1.
Texture data on disk is cached into system memory in an image cache’s mem region.
2. Texture data in the tex region part of the mem region is used to update texture
memory.
007-1680-100
543
15: ClipTextures
Image cache
Texture data on disk
Tex region
Image cache grid
Texture memory
Figure 15-5
Cliptexture Cache Hierarchy
Toroidal Loading
In order to minimize the bandwidth required to download texels from system to texture
memory, the image cache’s tex regions are updated using toroidal loading. A toroidal
load assumes that changes in the contents of the clip region are incremental, such that the
update consists of:
•
New texels that need to be loaded.
•
Texels that are no longer valid.
•
Texels that are still in the clip region, but have shifted position.
Toroidal loading minimizes texture downloading by only updating the part of the
texture region that needs new texels. Shifting texels that remain visible is not necessary,
since the coordinates of the clip region wrap around to the opposite side.
Invalid Borders
Being able to impose alignment requirements to the regions being downloaded to texture
memory improves performance. Cliptextures support the concept of an invalid border to
provide this feature. It is the area around the perimeter of a clip region that can’t be used.
The invalid border shrinks the usable area of the clip region, and can be used to
dynamically change the effective size of the clip region. Shrinking the effective clip size
can be a useful load control technique.
544
007-1680-100
Overview
When texturing requires texels from a portion of an invalid border at a given MIPmap
level, the texturing system moves down a level, and tries again. It keeps going down to
coarser levels until it finds texels at the proper coordinates that are not in the invalid
region. This is always guaranteed to happen, since each level covers the same area with
less texels (coarser level texels cover more area on textured geometry). Even if the
required texel is clipped out of every clipped level, the unclipped pyramid levels will
contain it.
You can use an invalid border to force the use of lower levels of the MIPmap to do the
following:
•
Reduce the abrupt discontinuity between MIPmap levels if the clip region is small:
using coarser LODs blends MIPmap levels over a larger textured region.
•
Improve performance when a texture must be roamed very quickly.
Since the invalid border can be adjusted dynamically, it can reduce the texture and
system memory loading requirements at the expense of a blurrier image.
Required texel
Clip center
Clip region
Fine
Invalid border
Required texel
Clip center
Coarser
Figure 15-6
007-1680-100
Invalid Border
545
15: ClipTextures
Updating the Clipcenter
To figure out what part of the texture must be loaded in each of the clipped levels, you
must know where the viewer is relative to the geometry being textured. Often this
position is computed by finding the location of the cliptextured geometry that is closest
to the viewer, and converting that to a location on the texture. This position is called the
cliptexture center and it must be updated every frame as the viewer moves relative to the
cliptextured geometry.
Centered
Center moves
Toroidal loads
Figure 15-7
Texture coordinates wrap
Same as centered
Clipcenter Moving
The clipcenter is set by the application for level 0, The cliptexture code then derives the
clipcenter location on all MIPmap levels. As the viewer roams over a cliptexture, the
centers of each MIPmap level move at a different rate. For example, moving the
clipcenter one unit corresponds to the center moving one half that distance in each
dimension in the next-coarser MIPmap level.
Most of the work of cliptexturing is updating the center properly and updating the
texture data in the clipped levels reliably and efficiently each frame.
Virtual Cliptextures on InfiniteReality Systems
Cliptextures save texture memory by limiting the extent of texture levels. Every level in
the mipmap is represented in texture memory, and can be accessed as the geometry is
546
007-1680-100
Overview
textured. On InfiniteReality systems, there are limits to the number of levels the
cliptexturing hardware can access while rendering, which restricts the cliptextures
maximum size.
This limit can be exceeded by only accessing a subset of all the MIPmap’s levels in texture
memory on each piece of geometry, “virtualizing” the cliptexture. The virtual offset is
sets a virtual “level 0” in the MIPmap, while the number of effective levels indicates how
many levels starting from the new level 0 can be accessed. The minlod and maxlod
parameters are used to ensure that only valid levels are displayed. The application
typically divides the cliptextured terrain into pieces, using the relative position of the
viewer and the terrain to update the parameter values as each piece is traversed.
Callback
Effective
levels
Callback
Effective
levels
Figure 15-8
Virtual Cliptexture Concepts
For more information about virtual cliptextures, see “Virtual ClipTextures” on page 593.
007-1680-100
547
15: ClipTextures
Cliptexture Support Requirements
Ideally, pfClipTextures would be interchangeable with pfTextures in OpenGL Performer.
Unfortunately, this is only partially true. The following sections describe some of the
differences between OpenGL Performer textures and cliptextures.
Centering
Every level is complete in a regular texture. Cliptextures have clipped levels, where only
the portion of the level near the cliptexture center is complete. In order to look correct, a
cliptextures center must be updated as the channel’s viewport moves relative to the
cliptextured geometry.
Cliptextures require functionality that recalculates the center position whenever the
viewer moves (essentially each frame). This means that a relationship has to exist
between the cliptexture and a channel.
Applying
Textures only need to be applied once. Cliptextures must be applied every time the center
moves (essentially each frame). In order to apply at the right time, cliptextures need to be
connected to a pfPipe.
Texel Data
A texture does not know where its data comes from. The application just supplies it as a
pointer to a region of system memory when the texture is applied.
Cliptextures need to update their contents as the center moves and they are reapplied
each frame. As a result, they need to know where their image data resides on the disk. In
order to maximize performance, cliptextures also cache their texel data in system
memory. As a result, cliptextures are a lot more work to configure, since you have to tell
them how to find their data on disk, and how you want the data cached in system
memory.
Special Features
Since cliptexture levels are so large, OpenGL Performer offers additional features not
available to regular textures.
548
007-1680-100
Overview
Insets
With certain restrictions, cliptexture levels can be partially populated, containing
“islands” of high resolution data. This can be useful if the application only needs
high-resolution texel data in relatively small, widely scattered areas of a large cliptexture.
An example of this might be an airline flight simulator, where high resolution data is only
needed in the vicinity of the airports used by the simulator.
For more information about insets, see “Cliptexture Insets” on page 601.
Virtualization on InfiniteReality Systems
To further increase the size of cliptextures that OpenGL Performer can use, the levels
themselves can be virtualized; It then selects a subset of all the available texture levels to
be loaded into memory. This requires additional support by the application. Virtual
cliptextures are described in detail in “Virtual ClipTextures” on page 593.
Multiple Pipe Support
Since cliptextures require both system and texture memory resources, OpenGL
Performer has provided functionality to share the system memory resources when a
cliptexture is used in a multipipe application. “Slave” cliptextures and a “master”
cliptexture share system memory resources, but have their own classes and texture
memory.
How Cliptextures Interact with the Rest of the System
As a result of their special requirements, cliptextures are used differently than pfTextures
with many different OpenGL Performer classes. The following sections describe these
differences.
Geostates
When everything is configured properly, a pfClipTexture is interchangeable with a
pfTexture when used in a geostate. Note that when using emulated cliptextures, the
pfClipTexture must be assigned to the pfGeoState before the pfGeoState is associated
with any pfGeoSet.
007-1680-100
549
15: ClipTextures
Pipes
A pfClipTexture can be connected to a pfMPClipTexture, a multiprocessing component,
which is connected to a pfPipe. From the pipe’s point of view, a pfMPClipTexture is
something it can apply to.
Channels
Some functionality must be supplied to update a cliptexture’s center as the channel
moves with respect to the cliptextured geometry. This functionality can be supplied by
the application, or OpenGL Performer can do it automatically if the application uses
clipcenter nodes.
A clipcenter node is added to the scenegraph and is traversed by the APP process just
like every other node in the scenegraph. When the clipcenter node is traversed by a
channel, the clipcenter node computes the relationship between the cliptextured
geometry and the channel’s eyepoint, and updates the cliptexture’s center appropriately.
Cliptexture Support in OpenGL Performer
Cliptexture is a large and diverse piece of functionality. As a result, cliptexture support
is found in nearly every major library in OpenGL Performer.
libpr Support
The pfImageCache class defines image caches which manage the updating of clipped
levels, pfImageTile classes are used to define non-clipped cliptexture levels and define
pieces of clipped levels downloaded from disk to system memory. The pfQueue class
supports read queues, which manage the read requests from disk to system memory in
image caches, while the pfClipTexture class itself defines cliptextures themselves, virtual
mipmaps composed of image caches and image tile levels. The pfTexLoad class defines
download requests when image caches download texels from system to texture memory.
libpf Support
The libpf library adds multiprocessing support for using cliptextures in scene graphs.
the pfMPClipTexture class ties together pfClipTextures, pfPipes, cliptexture centering
functionality (often pfuClipCenterNode nodes) and the application itself in a
550
007-1680-100
Overview
multiprocessing environment. additional functionality in the pfPipe class ensures that
cliptextures are applied properly.
libpfutil Support
The libpfutil library provides easy to use clipcentering functionality through the
pfuClipCenterNode class, a subclass of the pfGroup class. This library also provides
traversals to simplify the work of finding cliptextures in a scene graph using
pfuFindClipTextures(), code for post loader configuration, where pfMPClipTextures are
created, and attached to pipes and clipcenter nodes using pfuProcessClipCenters() and
pfuProcessClipCentersWithChannel(). The pfuAddMPClipTextureToPipes() and
pfuAddMPClipTexturesToPipes() routines connect pfMPClipTextures to the proper
pipes, handling multipipe issues in a clean way. Load time configuration is simplified
using the pfuInitClipTexConfig(), pfuMakeClipTexture(), and
pfuFreeClipTexConfig() along with the appropriate callbacks for image caches and
image tiles. Image cache configuration is supported with pfuInitImgCacheConfig(),
pfuMakeImageCache(), and pfuFreeImgCacheConfig() routines.
libpfdu Support
The cliptexture configuration file parsers are supported here; pfdLoadClipTexture() and
pfdLoadClipTextureState() work with cliptexture configuration files to simplify the
creation and configuration of cliptextures. The companion programs that create and
configure pfdLoadImageCache() and pfdLoadImageCacheState(). All of these parsers
use the pfuMakeClipTexture() and pfuMakeImageCache() configuration routines.
libpfdb Support
Example cliptexture loaders, including the libpfim example cliptexture loader, the
libpfct demo loader, the libpfspherepatch loader, and the libpfvct virtual
pseudo loader are all included here.
Sample Applications
Performer also ships with examples of applications that support cliptexturing. The main
example is clipfly (a modified version of perfly). Another program that uses
cliptextures is the clipdemo sample.
007-1680-100
551
15: ClipTextures
Cliptexture Manipulation
While the scene graph is being viewed, the application may want to dynamically alter
the appearance or performance characteristics of the cliptexture. The mpcliptexture
provides functionality to support parameter changes in the APP process, providing
frame-accurate updating. Here are some of the parameters that might be changed.
Load Control
The DTR functionality (described in detail elsewhere in this chapter) is largely automatic.
Some high performance applications may need to adjust DTR parameters to improve
appearance performance trade-offs.
Invalid Border
The invalid border can be adjusted at runtime to shrink the effective size of the clip
region. This might be done to provide additional load control beyond the per-level
control that DTR provides.
Share Masks
When operating master and slave cliptextures in a multipipe application, the application
may want to change the sharemask, which controls the synchronization of parameters
between master and slave cliptextures.
Read Function
The image cache creates requests to read image tiles from disk to the image cache’s
system memory cache. The read function processes these requests and actually does the
data transfer. OpenGL Performer provides set of read functions that attempts to do
direct-IO reads for speed, but falls back to normal reads if direct IO is not possible.
The application can replace the OpenGL Performer default function with its own custom
read function. This could be useful for implementing special functionality, such as
dynamic decompression pfClipTexture data.
552
007-1680-100
Cliptexture API
Read Queue Sorting
The read queue provides dynamic sorting of the read requests to improve performance
and minimize latency. The application can provide custom sorting routines.
Cliptexture API
Cliptexturing has a large API. Not only is there are lot of cliptexture functionality
scattered throughout the library, but there is often more than one way to use a particular
piece of functionality. In order to make things clearer, and make it easier to use the API
described here, the cliptexture API is grouped and ordered in the same way an
application writer would use it.
The API is grouped into four sections:
•
Preprocessing the cliptexture data.
•
Configuring cliptextures and image caches.
•
Post-load-time configuration.
•
Run-time manipulation.
Preprocessing ClipTextures
Before using cliptextures, large textures must be preprocessed, as follows:
1.
Start with the highest-resolution version of the image (texture) and build a MIPmap
of the image.
2. Choose a clip size.
3. Tile each MIPmap level.
Every image that is larger than the clip size must be cut into tiles. All of the tiles in
one MIPmap level must be equal in size. You generally choose a tile size that is
about 1/4 of the clip size or less.
4. Divide the levels into separate files to maximize download performance.
The files should be named properly so that the image caches can access them.
007-1680-100
553
15: ClipTextures
5. If the configuration parsers are used, cliptexture configuration files are also created
at this time.
The following sections describe the steps in this procedure in greater detail.
Building a MIPmap
Building a MIPmap of an image requires an algorithm that performs the following tasks:
1.
Start with the highest-resolution version of the image (texture). The image
dimensions in pixels must be in powers of 2, for example, 8192 X 8192.
2. Average every four adjacent texels of a high resolution image into a single texture
(essentially blurring it and shrinking it by a factor of two in both dimensions).
3. Save the result as a new, blurrier, smaller image.
4. Convert the MIPmaps into a compatible format.
5. Repeat the first two steps with each blurrier image until you have a single texel
whose color is the average of all the texel colors in the original image.
Each successive reduction is called a level of detail (LOD). The more the reduction, the
higher the level of detail, the coarser the image.
There are a variety of tools that tile textures. OpenGL Performer provides some simple
ones available in the /usr/share/Performer/src/tools directory for IRIX and
Linux and in %PFROOT%\Src\tools for Microsoft Windows. They are listed in
Table 15-1.
Table 15-1
554
Tiling Algorithms
Program
Description
rsets
Shrinks and tiles one or more .rgb image files recursively. rsets stops tiling when it
reaches the clip size you give it. rsets assumes that the original image is square.
rgb2raw
Converts .rgb images into a raw format that can be downloaded directly into texture
memory. Files should be in a raw format to avoid conversions at download time.
shrink
Is a subset of rsets functionality; makes a tree-like structure of LOD images from
an .rgb image.
to5551
Converts from .rgb to the 5551 raw format.
007-1680-100
Preprocessing ClipTextures
Table 15-1
Tiling Algorithms (continued)
Program
Description
to888
Converts from .rgb to the 888 raw format.
to8888
Converts from .rgb to the 8888 raw format.
viewtile Enables you to view a raw format image tile.
For more information about MIPmaps, see the OpenGL Programming Guide.
Formatting Image Data
The texel data must be in a format that can be used in OpenGL Performer textures. This
means the texels must have contiguous color components, whose size and type match a
supported format. Keep in mind that these texels will be loaded dynamically, on an
as-needed basis, so the smaller the size of each texel, the better the performance of the
cliptexture. You should choose the smallest texel format that provides acceptable color
quality. A good choice might be RGBA 5551, which takes up 16 bits per texel.
OpenGL Performer provides some tools for converting from rgb format to 5551 or 888
RGBA. They are named to5551 and to888 and are found in
/usr/share/Performer/src/tools for IRIX and Linux and in
%PFROOT%\Src\tools for Microsoft Windows.
For more information about file formats, see “Building a MIPmap” on page 554.
Tiling an Image
Dividing a texture into tiles allows you to look at a subset of all texels in the texture. In
this way, you can selectively download from disk into the texture memory only those
texels that the user is viewing and those they might soon look at. Since downloading
texture tile files from disk to texture memory takes a long time, the image cashes decide
which tiles a viewer might need next and download them in advance.
Note: In the highest resolution LOD, one texel corresponds to one pixel.
007-1680-100
555
15: ClipTextures
Texel tiles in each level are loaded into memory separately, from coarsest to finest. The
high-resolution tiles take longer to download than the coarser tiles. If a viewer advances
through a scene so quickly that the high-resolution tiles cannot download from disk into
texture memory in time, lower-resolution tiles are displayed instead. The effect is that if
the viewer goes too fast, the tiles become blurry. When the viewer slows down, the tiles
displayed are less coarse.
Using lower instead of higher-resolution levels is controlled by cliptexture’s load control
mechanism, DTR. Without DTR, OpenGL Performer waits for all of the levels to
download before displaying any one of them. DTR removes this restriction, displaying
the levels that have been downloaded.
If you want to break up a .rgb image into tiles, OpenGL Performer provides the subimg
program in /usr/share/Performer/src/tools for IRIX and Linux and in
%PFROOT%\Src\tools for Microsoft Windows.
Tile Size
Small tiles, while less efficient, are better at load leveling, since the time it takes to load a
new tile into system memory is smaller. It also means that the total size of an image cache
in system memory can be smaller. We’ve found that tile sizes of 512 x 512 and 1024 x 1024
provide a good trade-off between download efficiency and low latency, but download
performance is very sensitive to system configuration. Experimenting is the best way to
find a good tile size.
Cliptexture Configuration
After preprocessing the texture data, you need to configure cliptextures. Configuration
is actually a two step process; the configuration that can be done by the scenegraph
loader, and the configuration that requires pfPipes and pfChannels to be present. This
section describes the first stage of configuration.
Configuration Considerations
An application must configure the cliptexture in two steps:
556
•
Loader—when the scene graph is constructed.
•
Post-loading—when the channel and pipes are known to the application.
007-1680-100
Configuration API
This process is complex. OpenGL Performer supplies a number of utilities to make the
job easier.
To manipulate cliptexture parameters, the application makes calls to the
pfMPClipTexture in the APP process. The pfMPClipTexture updates the cliptexture in a
frame-accurate manner.
Load-Time Configuration
This is the time the scene graph is being constructed. Geostates are pointed to
cliptextures; the cliptextures themselves are created and configured using the cliptexture
configuration files and the libpfdu parsers. If the application does its own
configuration, it should use the libpfutil routines to simplify the process and to
ensure adequate error checking. If the application opts to use OpenGL Performer
clipcentering support, clipcenter nodes are inserted into the scene graph at the root of the
cliptextured geometry and connected to the corresponding cliptexture.
Note: When using emulated cliptextures, assign the cliptexture to the pfGeoState first,
then assign the pfGeoState to pfGeoSets.
Post-Load-Time Configuration
At this stage the scene graph has been created and the channels and pipes have been
defined. The libpfutil traversers (pfuProcessClipCenters() or
pfuProcessClipCentersWithChannel()) are used to create pfMPClipTextures,
connecting them with the appropriate cliptextures and clipcenter nodes. These routines
return a list of pfMPClipTextures, which can be used with
pfuAddMPClipTextureToPipes() and pfuAddMPClipTexturesToPipes() to attach the
pfMPClipTextures to the appropriate pfPipes. These routines can be used for single pipe
and multipipe applications with little or no change to the calling sequence.
Configuration API
Since cliptexture configuration is complex, we provide three different cliptexture
configuration API layers, allowing different trade-offs between flexibility and simplicity.
007-1680-100
557
15: ClipTextures
libpr Functionality
The lowest layer, using the libpr calls, is the most complex and difficult. A cliptexture
has the same configuration requirements as a MIPmapped pfTexture, where texel format,
type and texture dimensions must be configured. In addition, cliptextures have to know
about system memory caching, the file configuration of the texture data, load control,
read queue, and other cliptexture specific configurations.
Using the libpr layer directly is not recommended, since it is error prone and does not
buy much flexibility compared to the libpfutil configuration routines. In the
following subsections are the libpr calls you must consider when configuring a
cliptexture directly.
These are the functions needed to configure the cliptexture itself. The cliptexture contains
two types of levels: image cache levels and image tile levels. Image caches support
clipped levels in a cliptexture. They know where their texture data resides on disk, they
understand clip regions, and set up system memory caching and updating. Every
properly-configured image cache points to an image tile, called a proto tile, which
contains global information about the texel format, size, and file information about the
image tiles the image cache uses to update clipped texture levels.
Configuring an Image Cache Level
Image tiles can be used by themselves to represent unclipped levels. Essentially the
unclipped level is represented by a single tile covering the entire level. Because image
tiles do not understand clip regions and cannot do dynamic updating, image tiles cannot
be used to represent clipped levels.
To configure an image cache level, use the following calls:
558
•
pfNewClipTexture()
•
pfTexName()
•
pfClipTextureVirtualSize()
•
pfClipTextureClipSize()
•
pfTexImage()
•
pfTexFormat()
•
pfClipTextureInvalidBorder()
•
pfClipTextureEffectiveLevels()
007-1680-100
Configuration API
•
pfClipTextureAllocatedLevels()
•
pfClipTextureLevel()
Configuring an Image Cache Proto Tile
There are also image tile calls in this sequence. They are used to configure the image
cache’s proto tile, which is used as a template for the tiles the image cache will use to load
texel data from disk to system memory cache.
To configure an image cache proto tile, use the following calls:
•
pfNewImageTile()
•
pfImageTileReadFunc()
•
pfGetImageTileMemInfo (page size)
•
pfImageTileMemInfo()
•
pfImageTileReadQueue()
•
pfImageTileHeaderOffset()
•
pfImageTileNumFileTiles()
•
pfImageTileSize()
•
pfImageTileFileName()
•
pfImageTileFileImageType()
•
pfImageTileMemImageType()
Configuring an Image Cache
To configure an image cache, use the following calls:
007-1680-100
•
pfImageCacheName()
•
pfImageCacheTexRegionOrigin()
•
pfImageCacheMemRegionOrigin()
•
pfImageCacheImageSize()
•
pfImageCacheMemRegionSize()
•
pfImageCacheTileFileNameFormat()
559
15: ClipTextures
•
pfImageCacheTexRegionSize()
•
pfImageCacheMemRegionSize()
•
pfImageCacheTex()
•
pfImageCacheTexSize()
•
pfImageCacheFileStreamServer()
•
pfImageCacheProtoTile()—Copies the information into the image cache’s proto
tile.
•
pfDelete (tmp_proto_tile)—Now that it is copied into the image cache, you can
delete it.
Configuring a pfTexture
Image caches can be used independently of cliptextures, if they are, they need to be
associated with a pfTexture, and that texture needed to be configured.
To configure a pfTexture, use the following calls:
•
pfTexImage()
•
pfTexFormat()
Configuring the Default Tile
Image caches can have a default tile defined, which is the tile to use if a tile on disk can’t
be found. Default tiles can be useful for “filling in” border regions of a cliptexture level.
Default tiles are covered in more detail in section “default_tile” on page 572.
To configure the default tile, use the following calls:
560
•
pfNewImageTile()
•
pfCopy() (proto to default)
•
pfImageTileFileName()
•
pfImageTileReadQueue()
•
pfImageTileDefaultTile()
007-1680-100
Configuration API
Configuring Image Tiles
Image tiles need their own configuration, since they need to know about the file they
should load from texel formats, etc.
To configure an image tile, use the following calls:
•
pfNewImageTile()
•
pfImageTileMemImageFormat()
•
pfImageTileFileImageFormat()
•
pfImageTileMemImageType()
•
pfImageTileSize()
•
pfImageTileHeaderOffset()
•
pfClipTextureLevel()
•
pfLoadImageTile()
Configuration Utilities
Using the libpr calls to configure a cliptexture is difficult and error prone. OpenGL
Performer provides utilities to make cliptexture configuration easier and more robust.
The configuration utility API is broken into two groups. One group is used to configure
cliptextures, the other configures image caches. Each group contains three functions, an
init function, a config function, and a free function. These functions work with a structure
that the application fills in.
The initialize function initializes the optional fields in the structure with default values,
and the mandatory fields with invalid values. Configuring the structure allows the
configuration function to do more error checking, and to allow the application to avoid
the tedium of filling in optional field. The application then sets fields in the structure to
parameterize how the cliptexture or image cache should be configured. The application
then calls the configuration function on the filled in structure. The free function is then
called with the structure to ensure that all allocated values are freed.
Cliptexture Configuration
Methods to configure the cliptexture include the following:
007-1680-100
561
15: ClipTextures
pfuInitClipTexConfig(pfuClipTexConfig *config)
Initialize the values of the pfuClipTexConfig structure that has been
allocated by the application.
pfuMakeClipTexture(pfuClipTexConfig *config)
Return a cliptexture configured as directed by the values in the
pfuClipTexConfig structure.
pfuFreeClipTexConfig(pfuClipTexConfig *config)
Free any malloc’d structures that the application or the initialize
function may have created.
Image Cache Configuration
Methods to configure the image cache include the following:
pfuInitImgCacheConfig(pfuImgCacheConfig *config)
Initialize the values of the pfuClipTexConfig structure that has been
allocated by the application.
pfuMakeImageCache(pfuImgCacheConfig *config)
Return a cliptexture configured as directed by the values in the
pfuClipTexConfig structure.
pfuFreeImgCacheConfig(pfuImgCacheConfig *config)
Free any malloc’d structures that the application or the init() function
may have created.
All of these functions are defined in libpfutil/cliptexture.c. The structures
themselves are defined in pfutil.h.
Filling in the Structures
Filling the pfuImgCacheConfig structure to create and configure the image cache is
considerably simpler than setting fields in the pfuClipTexConfig structure. This is
because the cliptexture configuration must also create and configure image cache and
image tiles to populate its levels. The configuration code does this supplying a function
pointer to configure the image cache levels and a function pointer for configuring image
tile levels. Each function pointer also has a void data pointer so you can pass data to the
functions. The function pointers expect functions with the following forms:
pfImageCache *exampleICacheConfigFunction(pfClipTexture *ct,
int level, struct _pfuCilpTexConfig *icInfo)
562
007-1680-100
Configuration API
pfImageTile *exampleITileConfigFunction(pfClipTexture *ct,
int level, struct _pfuClipTexConfig *icInfo)
The cliptexture and image cache configuration parsers, described in the next section, use
the configuration utilities. You can look at the parsers as example code. For example, you
may want to look at pfdLoadImageTileFormat() and pfdLoadImageCache() formats for
example functions for the function pointers. The parsers are in the
/usr/share/Performer/src/lib/libpfdu/pfdLoadImage.c file for IRIX and
Linux and in file %PFROOT%\Src\lib\libpfdu\pfdLoadImage.c for Microsoft
Windows.
Configuration Files
The easiest and most commonly used method to configure cliptextures is to create
cliptexture and image cache configuration files, then use the configuration parsers to
create and configure cliptextures. The configuration files can be created and stored along
with the texture data files. Configuration files allow an application or loader to simply
call a single function to create and configure cliptextures.
Configuration files are ascii text files containing a token parameter format. Values are
separated by white space and the token parameter sequences can be placed in the file in
arbitrary order. Comments can also be added to the configuration files, making them
self-documenting.
Using Configuration Files
Four parser functions are available to create and configure cliptextures and image caches
using configuration files:
pfClipTexture *pfdLoadClipTexture(const char *fileName)
pfImageCache *pfdLoadImageCache(const char *fileName)
These parser functions take a configuration file name, and use it to configure and create
a cliptexture or an image cache respectively. The cliptexture configuration file may refer
to image cache configuration files, which will be searched for and used automatically.
Two other versions of these parsers also take a pointer to a configuration utility structure.
This allows you to preconfigure using the configuration structure and then finish with
the parser and configuration files.
007-1680-100
563
15: ClipTextures
pfClipTexture *pfdLoadClipTextureState(const char *fileName,
pfuClipTexConfig *state)
pfImageCache *pfdLoadImageCacheState(const char *fileName,
pfuImgCacheConfig *state)
The parsers use OpenGL Performer’s pfFindFile() functionality to search for the
configuration files. The parsers support environment variable expansion and relative
pathnames to make it simpler to create configuration files that refer to other
configuration or data files.
Creating Configuration Files
To successfully use cliptextures, you must first prepare the texture data and create the
configuration files:
1.
Create an image cache configuration file for each level using an image cache in the
cliptexture.
The configuration file should describe the following:
•
Format and tiling of the texture data.
•
Location and names of the files containing the texture data.
•
Size of the tex region in texture memory.
•
Size and layout of the mem region in system memory.
2. Create a cliptexture configuration file.
It contains the following:
564
•
Name and location of each image cache configuration file.
•
Names and locations of the texture data for each image tile level in the
cliptexture. Remember, image tile levels cannot be clipped levels; so, they can
only be used in the pyramid levels. Image cache levels can be used anywhere.
•
General properties of the cliptexture.
•
Look at the example cliptexture configuration files in the
/usr/share/Performer/data/clipdata directory for IRIX and Linux and
in %PFROOT%\Data\clipdata for Microsoft Windows. The cliptexture
configuration files use the .ct suffix. The image cache configuration files use .ic
for their suffixes.
007-1680-100
Configuration API
3. Test the image cache configuration files individually, using the
pguide/libpr/C/icache program.
4. Test the cliptexture configuration file using the /sample/C/clipfly or
/sample/C++/clipdemo sample appllications or the
/pguide/libpr/C/cliptex or the /pguide/libpf/C/cliptex programs.
5. When the configuration and data files are complete and tested, your application can
create and configure a cliptexture by calling pfdLoadClipTexture(fname) using the
name of the cliptexture configuration file. If more control is needed, you can use
pfdLoadClipTextureState(fname, state) initializing and configuring the
configuration utility cliptexture structure, pfuClipTexConfig.
Configuration File Tips
Unfortunately, cliptexture configuration is not trivial, even with documentation and
example programs. Success in creating working configuration files requires a two-prong
approach:
•
Keep them simple: set the minimum number of fields possible. Take advantage of
the default values. Try to find a similar example configuration file to copy from.
•
Work bottom-up: create and test image cache configuration files first, gradually
building up to the cliptexture configuration file.
We have found that parameterized naming of the image caches and tile files works the
best. If you have named your files consistently, this can be easy. If things do not work,
you can fall back and name your file explicitly as a sanity check. Read the error messages
carefully; they try to point out where in the configuration file the parser found problems.
If you need more information, try rerunning the program with PFNFYLEVEL set to 5 or
9.
A number of example configuration files and cliptextures are available on the OpenGL
Performer release. Working from one of them can save a lot of time. Some places to look
are the following:
•
data/clipdata/hunter
•
data/clipdata/moffett
•
data/asddata
Note that emulated cliptextures are currently incompatible with pfASD.
007-1680-100
565
15: ClipTextures
Cliptexture Loaders
Finally, your application might be able to take advantage of some of the cliptexture
loaders. The libpfim loader supports loading a cliptexture, and updating its center as a
function of viewposition. The libpfct loader creates a cliptexture with simple terrain.
Virtual cliptextures, mentioned in “Virtual ClipTextures” on page 593, can also be created
using the libpfspherepatch or libpfvct loaders. These loaders can be used as
examples if you need to write your own loader that supports cliptextures. The libpfct
and the libpfspherepatch loaders are the most up-to-date cliptexture loaders and
include support for emulated cliptextures as well as support for virtual cliptextures on
InfiniteReality systems.
Image Cache Configuration File Details
Image cache configuration files supply the following information to OpenGL Performer:
•
Format of the texel data.
•
Size of the entire texture at a particular MIPmap level.
•
How to find the files containing the texel data for this image cache.
•
Size and layout of image cache tiles in memory.
•
Size of the image cache that should be kept in texture memory.
•
A default image tile to use if one is missing.
•
The size each level should be clipped to.
•
The amount of border that should be invalidated at each level.
Configuration Fields
Configuration fields are either tokens or parameter values, as listed in Table 15-2. All
fields are character strings and all parameters must be separated by white space. The
token names marked with an asterisk (*) are optional and default to reasonable values.
Table 15-2
566
Image Cache Configuration File Fields
Token Name
Parameters
Description
ic_version2.0
no data field
Start of image cache config files: type and version
*tex_size
3 integers
Area of tex memory for level if not tex_region_size
*header_offset
integer
Beginning of file to skip over in bytes
007-1680-100
Configuration API
Table 15-2
Image Cache Configuration File Fields (continued)
Token Name
Parameters
Description
*tiles_in_file
3 integers
Dimensions of grid of tiles stored in each file
*s_streams
filepath list
List of streams used to access files in S dimension
*t_streams
filepath list
List of streams used to access files in T dimension
*r_streams
filepath list
List of streams used to access files in R dimensions
*default_tile
filepath string
Tile to use if expected tile is not available
*page_size
integer
System page size; memory allocation alignment
*read_func
1 or 2 strings
Custom read function; library, func or func in app
*lookahead
integer
Extra tiles in mem region for lookahead caching
ext_format
string
External format of stored texels
int_format
string
Internal format used by graphics hw
img_format
string
Image format of stored texels
icache_size
3 integers
Size of complete image level in texels
tex_region_size
3 integers
Area to load in texture memory; matches clip size
mem_region_size
3 integers
Dimensions of system memory cache in tiles
tile_size
3 integers
Dimensions of each file in texels
tile_format
scanf-style string
Parameterized path to tile files
tile_params
list of symbols
Parameter types in order in tile_format string
Image Cache Configuration File Description
The ic_version2.0 token must be first in an image cache configuration file. This token
identifies the file as an image cache configuration file and the format (version) of the
configuration file.
Next the parser looks for tokens and any associated data values. In general, the order of
the tokens in the file must follow the sequence specified in the table above. The tokens
marked with an asterisk are optional. Optional tokens have default values, which are
used if the token and value are omitted.Tokens can have the following:
007-1680-100
567
15: ClipTextures
•
No arguments
•
A fixed number of arguments
•
A variable number of arguments
If a token has a fixed number of arguments, the token must be followed by a white
space-separated list containing the specified number of arguments. If the token has a
variable number of arguments, one of its arguments specifies the number of arguments
used.
Any time a token is expected by the parser, a comment can be substituted. A comment
cannot be put anywhere in the file, however. For example, if a token expects arguments,
you cannot place a comment between any of them; you have to place it after all of the
previous tokens arguments. There are a variety of supported comment tokens; they are
interchangeable. The comment tokens are #, //, ;, comment, or rem.
ext_format, int_format, and img_format
One of the first things that must be specified in an image cache is the format of the texel
data. This includes the external format (ext_format), internal format (int_format)
and image format (img_format). The arguments expected by these format parameters
are the ASCII string names of the format’s enumerates. For example, a valid external
format would be ext_format PFTEX_FLOAT. Consult the pfTexture man pages for a
list of the valid formats of each type.
icache_size, mem_region_size, and tex_region_size
The next set of parameters that must be specified in the image cache configuration file is
its size on disk, in system memory, and in texture memory. The icache_size token
requires the size of the image cache. This means the dimensions, in texels, in the s, t, and
r dimensions of the complete texture at this level. Since three dimensional textures are
not currently supported, the r parameter will always be 1.
An image cache’s texels are organized into a set of fixed sized pieces, called tiles. Both in
system memory and on disk, the texels are broken up this way. At any given time, an
array of these texel tiles are cached in system memory. They are arranged as an array in
system memory. If the center of the image cache nears the edge of this array, the most
distant tiles are dropped out, and new tiles are read in from disk. The larger the array of
tiles in system memory, the more of the complete texture is cached there, and the less
568
007-1680-100
Configuration API
likely new tiles may need to be swapped in. The benefit is offset by the cost of tying up
more system memory to hold the texel tiles.
The arrangement and dimensions of tiles in system is defined for each image cache, and
is set with the mem_region_size token. This token expects three arguments which
determine the number of tiles in the s, t, and r dimensions of the grid. Since three
dimensional textures aren’t currently supported, the r dimension is always 1.
A subset of the texels in system memory are cached in the texture memory itself. These
texels are arranged in a rectangular region. The dimensions of this region are defined by
the tex_region_size token. It expects three arguments, the number of texels in the s,
t, and r dimensions. Again, since three dimensional textures are not supported, the r
value is always 1.
The image cache configuration file allows some leeway in the arrangement of texel tiles
on disk. There can be one or more tiles on each disk file, and the file itself could contain
non-texel information at the beginning of the file. The tiles themselves can have
user-specified dimensions. While there is some flexibility in how tiles are stored in files
on disk, there are restrictions also. Any header must be the same size for every file in an
image cache. The same is true for the tile size, and the number and layout of tiles in each
file. If there is more than one tile in a file, the tiles must be arranged in row-major order.
In other words, as you pass from the first tile to the last, the s dimension must be
incrementing fastest.
tile_format and tile_params
The image cache texel data is stored in one or more files. The configuration file provides
a way for OpenGL Performer to find these files. The files usually have similar names,
varying in a predictable way, such as by tile position in the image cache array and size of
the image cache. The files themselves are grouped in on or more directories. The file
name and file path information is divided into a number of groups within the
configuration file. There is a scanf-style string specifying the path to find image cache
files. There are a number of parameters in the string that vary as a function of the tile
required and the characteristics of the image cache.
The next group of tokens describes the location of the configuration files defining the
location of the texture data tiles for the image cache. You can define the texture tile
configuration filenames with a scanf-style string containing parameter values, as is
done with image caches. To create parameterized image cache names, you must define
the tile_format and tile_params tokens.
007-1680-100
569
15: ClipTextures
The tile_format token is followed by a scanf-style string describing the file path and
filename of the image cache configuration files. The argument contains constant parts,
interspersed with %d or %s parameters. The number of parameters must match the
number of symbols supplied as parameters to the tile_params token. If the
tile_format string starts with the pattern $ENVNAME, ${ENVNAME}, or $(ENVNAME),
then the value of ENVNAME will be assumed to be an environment variable and expanded
into the base name.
The possible values of the image tile file name parameters is given in the table below.
Table 15-3
Image Tile Filename Tokens
Image Tile Filename Tokens
Description
PFIMAGECACHE_TILE_FILENAMEARG_VSIZE_S
Virtual size S width
PFIMAGECACHE_TILE_FILENAMEARG_VSIZE_T
Virtual size T width
PFIMAGECACHE_TILE_FILENAMEARG_VSIZE_R
Virtual size R width
PFIMAGECACHE_TILE_FILENAMEARG_TILENUM_S
Tiles from origin in S
PFIMAGECACHE_TILE_FILENAMEARG_TILENUM_T
Tiles from origin in T
PFIMAGECACHE_TILE_FILENAMEARG_TILENUM_R
Tiles from origin in R
PFIMAGECACHE_TILE_FILENAMEARG_TILEORG_S
Texels from origin in S
PFIMAGECACHE_TILE_FILENAMEARG_TILEORG_T
Texels from origin in T
PFIMAGECACHE_TILE_FILENAMEARG_TILEORG_R
Texels from origin in R
PFIMAGECACHE_TILE_FILENAMEARG_STREAMSERVERNAME
From streams
PFIMAGECACHE_TILE_FILENAMEARG_CACHENAME
The tile_base value
PFIMAGECACHE_TILE_FILENAMEARG_FILENUM_S
Files from origin in S
PFIMAGECACHE_TILE_FILENAMEARG_FILENUM_T
Files from origin in T
PFIMAGECACHE_TILE_FILENAMEARG_FILENUM_R
Files from origin in R
header_offset, tiles_in_file, and tile_size
The header_offset argument specifies the size of the file’s header in bytes. This many
bytes will be skipped over as a file is read. The tiles_in_file token requires three
arguments, specifying the number of tiles in the s, t, and r dimensions. The r dimension
570
007-1680-100
Configuration API
must always be 1, since 3D textures are not supported. The tile_size parameter
defines the texel dimensions of each tile in s, t, and r. Again, r must be 1. Both the
header_offset and the tiles_in_file tokens are optional. They default to the
values 0 and 1 1 1, respectively, specifying no header and a single tile in each file.
One of the major bottlenecks to sustained cliptexture performance is the speed of
copying texels from disk to system memory. Cliptextures can be configured to maximize
the bandwidth of this transfer by distributing image tiles over multiple disks and
downloading them in parallel. The streams section of the configuration file is used for
this purpose.
num_streams, s_streams, t_streams, and r_streams
A stream, short for stream device, can be thought of as a separate disk that can be
accessed in parallel with other disks. Each disk is mounted in a file system and, therefore,
has a unique filepath segment. The streams tokens allow you to identify these stream
filepath segments and how the image tiles are distributed among them. The stream
devices are arranged in a three dimensional grid with s, t, and r dimensions just like the
image tiles are in memory. The stream device is accessed by taking the position of the tile,
counting tiles from the origin in the s, t, and r directions, and generating a coordinate,
modulo the number of stream devices in the corresponding s, t, and r directions. The s, t,
and r values generated are used to look up the appropriate stream device. If the stream
server name is part of the tile file name format string, it effects which disk is used to find
the tile.
Stream servers improve bandwidth at the expense of duplicating image tiles over
multiple disks. You must insure that the proper image tiles are available for any disk
which is addressed by the tile’s s, t, and r coordinates modulo the available number of
stream servers for each of those dimensions. The stream server tokens are optional. The
s_streams token is followed by a list of filepaths. These are the names that will be
indexed from the list by taking the s coordinate of the tile’s position in the image cache
grid, modulo the number of s stream devices. The names in the s_stream list do not
have to be unique.
The t_streams and r_streams tokens work in exactly the same way, in the t and r
directions, respectively.
Sometimes only a subregion of the entire cliptexture is of interest to the application. This
is especially true when you consider that the number of tiles in the s, t, and r directions
must all be a power of two. To save space, improve performance, and make creating
image caches more convenient, a default tile can be defined, and tiles of no interest can
007-1680-100
571
15: ClipTextures
simply be omitted. If a tile cannot be found and a default tile is defined, then the default
one is used in place of the missing one.
default_tile
Unlike normal tiles, which are read from disk as they are needed, the default tile is loaded
as part of the configuration process. The tile is named in the configuration file as the
argument to the default_tile token. The argument is a filepath to the default tile. If
the tile_base token has been defined, it is pre-pended to the file path; otherwise, it is
used as is.
Cliptexture Configuration File Details
Image cache configuration files supply the following information to OpenGL Performer:
•
format of the texel data
•
Size of the highest resolution level (level 0)
•
Size of clipped levels in texture memory
•
How to find the configuration files for each image cache
•
Size of the smallest level to be loaded as an image cache
•
Number of effective levels for virtual cliptextures
•
Number of allocated levels for virtual cliptextures
Additionally, if no image-cache configurations are used, the cliptexture configuration file
will include the following:
572
•
Size of image tiles on disk
•
Number of tiles to be stored in RAM for each level
•
Format of tile filenames
•
Size of header in tile files
007-1680-100
Configuration API
Configuration Fields
Configuration fields are either tokens or parameter values, as listed in Table 15-4. All
fields are character strings and all parameters must be separated by white space.
Table 15-4
Cliptexture Configuration File Fields
Token Name
Parameters
# or // or ; or comment comment
007-1680-100
Description
comment symbols; comment to end of line
ct_version2.0
no data field
the beginning of the file: type and version
ext_format
string
external format of stored texels
int_format
string
internal format used by graphics hw
img_format
string
image format of stored texels
virt_size
3 integers
size of complete texture at level 0 (finest level)
clip_size
integer
size of clip region square for clipped levels
*invalid_border
integer
width of clip region perimeter to not use
*tile_size
3 integers
size of tiles (used if no icache config files)
*smallest_icache
3 integers
smallest icache-level dimensions
*lookahead
integer
extra tiles in mem region
*icache_format
scanf string
icache fnames: no field? list files
*effective_levels
integer
levels used for texturing in virtual cliptexture
*icache_params
string list
format tokens in order
*icache_files
list of filenames
only if icache_format is default
*tile_files
list of filenames
pyramid; only if tile_format default
*effective_levels
integer
levels used for texturing in virtual cliptexture
*allocated_levels
integer
total virtual cliptexture levels in texture memory
*header_offset
1 integer
byte offset to skip user’s file header
*tiles_in_file
3 integers
Image tile arrangement in each file
*read_func
1 or 2 strings
custom read function; lib & func or func in app
573
15: ClipTextures
Table 15-4
Cliptexture Configuration File Fields (continued)
Token Name
Parameters
Description
*tile_format
scanf string
Tile filename format
*tile_params
string list
format parameter tokens in order
*page_size
integer
system page size; memory allocation alignment
Cliptexture Configuration File Description
The ct_version2.0 token must be first in an cliptexture configuration file. This token
identifies the file as an cliptexture configuration file and the format (version) of the
configuration file.
Next the parser looks for tokens and any associated data values. In general, the order of
the tokens in the file must follow the sequence specified in the table above. The tokens
marked with an asterisk are optional. Optional tokens have default values, which are
used if the token and value are omitted.
Tokens can have the following:
•
No arguments
•
A fixed number of arguments
•
A variable number of arguments
If a token has a fixed number of arguments, the token must be followed by a white
space-separated list containing the specified number of arguments. If the token has a
variable number of arguments, one of its arguments specifies the number of arguments
used.
Any time a token is expected by the parser, a comment can be substituted. A comment
can’t be put anywhere in the file, however. For example, if a token expects arguments,
you can’t place a comment between any of them; you have to place it after all of the
previous tokens arguments. There are a variety of supported comment tokens; they are
interchangeable. The comment tokens are #, //, ;, comment, or rem.
574
007-1680-100
Configuration API
ext_format, int_format, and img_format
One of the first things that must be specified in a cliptexture is the format of the texel data.
This includes the external format (ext_format), internal format (int_format) and
image format (img_format). The arguments expected by these format parameters are
the ASCII string names of the format’s enumerates. For example, a valid external format
would be ext_format PFTEX_FLOAT. Consult the pfTexture man pages for a list of the
valid formats of each type.
virt_size and clip_size
The next group of tokens characterizes the image cache itself. The virt_size token
expects three integer arguments. They define the s, t, and r dimensions of the level 0 layer
of the cliptexture in texels. The clip_size token describes the size of each layer that
exists in texture memory. It takes one integer, describing the s and t dimensions of the
clipped region. This value is the same for all levels of a cliptexture. If the image cache
configuration files’ clip_size differs from this value, the cliptexture overrides it.
invalid_border
The invalid_border defines the region of each clipped level that should not be used.
If a texel is needed in that region, the next level down is used instead. If the invalid border
is large, the system may have to go down multiple levels, or even down to the pyramidal,
unclipped part of the MIPmap. The invalid border argument is a single integer,
describing the width of the border in texels.
smallest_icache
The smallest_icache token describes the s, t, and r dimensions of the lowest level
that is described as an image cache. This parameter is needed because the unclipped,
pyramidal part of the MIPmap can also be configured as image caches. This is an optional
token. If it is not included in the file, the last clipped level is considered the smallest
image cache in the cliptexture.
icache_files, icache_format and icache_params
The next group of tokens describes the location of the configuration files defining the
image cache levels of the cliptexture. There are two methods of describing where the
image cache configuration files. You can explicitly list the filenames in order with
icache_files.
007-1680-100
575
15: ClipTextures
The other method is to define the image cache configuration filenames with a scanf-style
string containing parameter values, as is done with image caches. This is usually the
preferred method. To create parameterized image cache names, you must define the
icache_format and icache_params tokens. If the format string starts with the
pattern $ENVNAME, ${ENVNAME} or $(ENVNAME), then the value of ENVNAME will be
assumed to be an environment variable and expanded into the base name.
The icache_format token is followed by a scanf-style string describing the file path
and filename of the image cache configuration files. The argument contains constant
parts, interspersed with %d or %s parameters. The number of parameters must match
the integer given with the num_icache_params token. The tile parameters themselves
follow the icache_params token.
icache_files
The number of parameters must match the number of parameters in icache_format.
All of these parameters are optional. The list of available parameter tokens is given in
Table 15-5.
Table 15-5
Parameter Tokens
Parameter Token Name
Description
PFCLIPTEX_FNAMEARG_LEVEL
Cliptexture level (top is 0)
PFCLIPTEX_FNAMEARG_LEVEL_SIZE
Largest value of level’s virtual size
PFCLIPTEX_FNAMEARG_IMAGE_CACHE_BASE
Value of icache_base
PFCLIPTEX_FNAMEARG_TILE_BASE
Value of tile_base
Uniquely naming that file for each level of the cliptexture, the parameter values are used
to construct the name of the image cache configuration file.
Near the bottom of the cliptexture, the size of lower levels are too small to warrant image
caches. These levels are specified directly, referring to a single filename containing a
single image tile for each level. The filenames for these tile files are specified in exactly
the same way as the image cache configuration files are. Instead of icache_base,
icache_format, num_icache_parameters, and icache_parameters,
tile_base, tile_format, num_tile_parameters, and tile_parameters are
used. The parameters available for use in the tile_format string are identical to the
ones used for icache_format.
576
007-1680-100
Configuration API
tile_files
If image cache configuration files and/or image tiles are to be explicitly named, they are
listed in order, from the top (largest) level to the bottom, using the icache_files and
tile_files tokens. These tokens can only be used if the corresponding format,
num_parameters, and parameter tokens are not. The number of filenames listed after
icache_files and tile_files must exactly match the number of cached and
uncached levels, respectively, in the cliptexture.
header_offset, tiles_in_file, and tile_size
The header_offset argument specifies the size of the file’s header in bytes. This many
bytes will be skipped over as a file is read. The tiles_in_file token requires three
arguments, specifying the number of tiles in the s, t, and r dimensions. The r dimension
must always be 1, since 3D cliptextures are not supported. The tile_size parameter
defines the texel dimensions of each tile in s, t, and r. Again, r must be 1. Both the
header_offset and the tiles_in_file tokens are optional. They default to the
values 0 and 1 1 1, respectively, specifying no header and a single tile in each file.
The image cache texel data is stored in one or more files. The configuration file provides
a way for OpenGL Performer to find these files. The files usually have similar names,
varying in a predictable way, such as by tile position in the image cache array and size of
the image cache. The files themselves are grouped in on or more directories. The file
name and file path information is divided into a number of groups within the
configuration file. There is a scanf-style string specifying the path to find image cache
files. There are a number of parameters in the string that vary as a function of the tile
required, and characteristics of the image cache.
tile_base, tile_format and tile_params
The tile_format token expects a scanf-style argument. If the string starts with the
pattern $ENVNAME, ${ENVNAME} or $(ENVNAME), then the value of ENVNAME will be
assumed to be an environment variable and expanded into the base name.
The argument contains constant parts, interpersed with %d or %s parameters.The tile
parameters themselves follow the tile_params token. The number of parameters must
match the number of parameters in tile_format.
007-1680-100
577
15: ClipTextures
The possible values of the image tile file name parameters are given in the table below.
Table 15-6
Image Tile Filename Tokens
Image Tile Filename Tokens
Description
PFIMAGECACHE_TILE_FILENAMEARG_VSIZE_S
Virtual size S width
PFIMAGECACHE_TILE_FILENAMEARG_VSIZE_T
Virtual size T width
PFIMAGECACHE_TILE_FILENAMEARG_VSIZE_R
Virtual size R width
PFIMAGECACHE_TILE_FILENAMEARG_TILENUM_S
Tiles from origin in S
PFIMAGECACHE_TILE_FILENAMEARG_TILENUM_T
Tiles from origin in T
PFIMAGECACHE_TILE_FILENAMEARG_TILENUM_R
Tiles from origin in R
PFIMAGECACHE_TILE_FILENAMEARG_TILEORG_S
Texels from origin in S
PFIMAGECACHE_TILE_FILENAMEARG_TILEORG_T
Texels from origin in T
PFIMAGECACHE_TILE_FILENAMEARG_TILEORG_R
Texels from origin in R
PFIMAGECACHE_TILE_FILENAMEARG_STREAMSERVERNAME
From streams
PFIMAGECACHE_TILE_FILENAMEARG_CACHENAME
The tile_base value
PFIMAGECACHE_TILE_FILENAMEARG_FILENUM_S
Files from origin in S
PFIMAGECACHE_TILE_FILENAMEARG_FILENUM_T
Files from origin in T
PFIMAGECACHE_TILE_FILENAMEARG_FILENUM_R
Files from origin in R
Optional Image Cache Configuration Files
If the cliptexture has a very regular structure from level to level, the cliptexture
configuration file can be augmented with some extra fields, and the image cache
configuration files dispensed with. We recommend you start with the image cache
configuration files, however, because it makes it easier to gradually create and test your
configuration files using the icache and cliptex utilities in the
/usr/share/Performer/src/pguide/libpr/C directory for IRIX and Linux and
in %PFROOT%\Src\pguide\libpr\C for Microsoft Windows.
Image cache configuration files can be removed if the image caches of the cliptexture are
essentially the same, and configuration of each image cache is simple. The image caches
should only differ in size between levels; the tile size, formats, tile filename format, etc.
578
007-1680-100
Post-Scene Graph Load Configuration
should be the same. Also image cache configuration files are not optional when features
like streams are configured.
To stop using image cache configuration files, you should add a tile_size token to the
cliptexture configuration file, and be sure to have tile_format and tile_params
specified.The tile specification in the cliptexture configuration file will be used for all tile
files: the ones used by the image caches and the ones representing pyramid levels.
In order to make the parser stop using the image cache configuration files, remove the
entries referring to them such as icache_format, icache_params, or
icache_tiles.
An example of a cliptexture configuration file that does not use image cache
configuration files is
/usr/share/Performer/data/clipdata/hunter/hl.noic.ct for IRIX and
Linux and %PFROOT%\Data\clipdata\hunter\hl.noic.ct for Microsoft
Windows.
Post-Scene Graph Load Configuration
There are a number of cliptexture configuration steps that cannot be completed until the
OpenGL Performer application’s pipes and channels have been created. This
configuration stage centers around configuring cliptextures to be properly applied and
centered each frame.
Two jobs must be accomplished. Each cliptexture must be attached to a pipe through its
own pfMPClipTexture so it can be applied each frame, and a centering callback must be
established to update the cliptexture as the channel’s viewpoint moves with respect to
the cliptextured geometry.
MPClipTextures
pfMPClipTexture is a multiprocess wrapper for a pfClipTexture. A pfMPClipTexture
allows you to do the following:
007-1680-100
•
Change the center of the pfClipTexture in the APP process.
•
Automatically schedule the necessary texture downloading (applying) in the CULL
process. Downloads are then performed in the DRAW process.
579
15: ClipTextures
•
Control the cliptexture parameters in the APP process.
APP
process
Clip Center
node
MPClipTexture
Pipe
Clip Texture
Figure 15-9
pfMPClipTexture Connections
Connecting MPcliptextures to pfPipes
To automatically apply of the pfClipTexture at the correct times and in the correct
processes, you must do the following:
1.
Create a pfMPClipTexture object.
2. Attach the pfMPClipTexture to the cliptexture you want to control.
3. Attach the pfMPClipTexture object to a pfPipe using the pfPipe.
Note: If you use pfMPClipTexture, you should never call either
pfUpdateMPClipTexture() or pfApplyMPClipTexture(); the pfPipe should do the
applying.
When you attach a pfMPClipTexture to a pfPipe using pfAddMPClipTextureToPipes()
or pfAddMPClipTexturesToPipes(), pfPipe automatically updates and applies
580
007-1680-100
Post-Scene Graph Load Configuration
pfClipTexture at the correct time. The functions take three arguments: a pfMPClipTexture
or list of pfMPClipTextures, a pipe to which to attach (called the master pipe), and a list
of other pipes the application wants to use with the pfMPClipTextures.
•
pfAddMPClipTextureToPipes(pfMPClipTexture, masterpipe, pipe_list)
•
pfAddMPClipTexturesToPipes(pfMPClipTexture_list, masterpipe, pipe_list)
The pipe_list is used for multipipe applications. It is the list of pipes that slave
pfMPClipTextures should be attached to. Setting pipe_list to NULL is equivalent to
adding slave pfMPClipTextures to every other pipe in the application.
There are additional libpf routines that can be useful:
•
pfRemoveMPClipTexture() detaches a pfMPClipTexture from a pfPipe. If a
pfMPClipTexture is removed that is the master of other pfMPClipTextures, the
slaves will be removed from their pipes as well.
•
pfGetNumMPClipTextures() returns the number of pfMPClipTextures attached to
a pfPipe.
•
pfGetMPClipTexture() returns a pointer to the pfMPClipTexture that is attached to
a pfPipe.
libpf Functionality
You can do this directly with the libpf API using the following calls:
•
pfNewMPClipTexture()—Create a new pfMPClipTexture.
•
pfMPClipTextureClipTexture()—Attach the pfMPClipTexture to the cliptexture.
•
pfAddMPClipTexture() - (a pfPipeCall)—Attach the pfMPClipTexture to a pipe.
•
pfMPClipTexturePipe()—Specifies to the pfMPClipTexture the pipe to which it is
attached.
pfMPClipTexture Utilities
OpenGL Performer provides utilities to make it easy to attach pfMPClipTextures to
pipes, and to automatically do pfMPClipTexture centering as well. As a bonus, the utility
code requires little or no changes to convert a single pipe application to a multipipe one.
007-1680-100
581
15: ClipTextures
To use the pfMPClipTexture utilities, you need to use OpenGL Performer’s clipcenter
nodes to center the pfMPClipTexture. clipcenter nodes are a subclass of pfGroup nodes.
They have additional functionality that allows them to connect to a pfMPClipTexture, the
cliptextured geometry (through their child nodes), and properly update the
pfMPClipTexture’s center each frame. At load time, clipcenter nodes are placed at the
root of the subtree containing the cliptextured geometry. All the cliptextures in the scene
are created configured and attached to the clipcenter node at this time as well.
Once you have a scenegraph with geometry, cliptextures, and clipcenter nodes, it is easy
to make pfMPClipTextures, attach them to pipes and to centering callbacks. The function
pfuProcessClipCenters() traverses the scene graph, looking for clipcenter nodes. As
each node is encountered, the function creates an MP cliptexture, attaches it to the
associated cliptexture and the clipcenter node, and saves a pointer to the MP cliptexture
in a pfList. When the function returns, it provides the list of MP cliptextures that were
created. The pfuProcessClipCentersWithChannel() routine performs the same
operations but also sets a channel pointer in the clipcenter node. When the channel
pointer is set, the clipcenter node only will update a pfMPClipTexture center when that
channel traverses it. This is useful for multichannel applications.
Clipcenter Node
In order for cliptextures to be rendered correctly, the clipcenter must move along with the
viewer. OpenGL Performer has made this task simpler by providing a special node for
the scene graph that does this calculation and applies it to the cliptexture each frame. This
node, called the clipcenter node, is a subclass of a pfGroup node. In addition to pfGroup
functionality, pfuClipCenterNode’s can do the following:
582
•
Points to the cliptexture. This allows cliptextures to be attached to clipcenter nodes
at load time.
•
Points to the geometry textured by the clipcenter node’s cliptexture. The clipcenter
node is assumed rooted in the subtree containing the cliptextured geometry.
•
Points to an optional simplified version of the cliptextured geometry to make
centering calculations go faster.
•
Points to the pfMPClipTexture attached to the cliptexture. The node also has API to
automatically create an pfMPClipTexture and attach it to the cliptexture.
•
Contains a replaceable post-APP callback function for updating a
pfMPClipTexture’s center.
•
Can point to a pfChannel and only update the pfMPClipTexture center when that
pfChannel traverses the clipcenter node.
007-1680-100
Post-Scene Graph Load Configuration
Channel
Clip Center
node
Scene graph
Geometry
MPClipTexture
Figure 15-10
pfuClipCenterNode Connections
The clipcenter node uses a simple algorithm, setting the cliptexture center to be the point
on the textured geometry closest to the viewer. Other algorithms can be used by
replacing the callback function.
Clipcenter nodes can be created by calling the utility routine pfuNewClipCenterNode().
There are set and get functions to attach cliptextures, channels, custom centering
callbacks, simplified cliptextured geometry, as well as a get to return the
pfMPClipTexture. See the pfuClipCenterNode man page for details on the API.
The clipcenter node source code is available in pfuClipCenterNode.C and
pfuClipCenterNode.h in the /usr/share/Performer/src/lib/libputil
directory for IRIX and Linux and in %PFROOT%\Src\lib\libpfutil for Microsoft
Windows. It is implemented as a C++ class with C++ and C API. It also has example code
illustrating how to subclass the clipcenternode further to customize it.
If the configuration has been done properly, and if pfuClipCenterNodes have been used
for centering, most of the per-frame operations for cliptextures is automatic. Centering is
computed and applied by the clipcenter nodes during the APP traversal, and cliptexture
application is automatically handled by the pfPipes attached to the pfMPClipTextures.
007-1680-100
583
15: ClipTextures
Using Cliptextures with Multiple Pipes
Cliptextures use a lot of texture memory, system memory (for their caches) and disk I/O
bandwidth. Many multipipe applications produce multiple views from the same
location, looking in different directions. It would be very inefficient to create a
completely separate cliptexture for each pipe; although there is separate texture memory
and graphics hardware from each pipe, the system memory and disk resources are
shared by the entire system.
Cliptextures have been designed to support multipipe rendering without excessive drain
on system memory and disk I/O bandwidth. Cliptextures that are to be used in multiple
pipes can be split into master and slave cliptextures. The master cliptexture is complete; it
contains an image cache and a region of texture memory to control. A slave cliptexture
points to its master and shares its image cache, using it to download into its own texture
memory. All the slave cliptextures share their master’s system memory cache and disk
I/O resources, reducing the load on the system.
Tex region
Mem region
Tex region
Tex region
Ma
ste
r
Sla
ve
Sla
Figure 15-11
ve
Master and Slave Cliptexture Resource Sharing
Making Masters and Slaves
Master and slave relationships can be established between image caches, cliptextures,
and pfMPClipTextures. The process starts with an object already configured the way you
want. Then another object of the same type is created and is set to be a slave of the
configured object. This is done with the setMaster() function. When an object is made the
slave of another object, it automatically configures itself to match it’s master. It also
makes all the connections necessary to share its master’s resources.
584
007-1680-100
Post-Scene Graph Load Configuration
If two cliptextures are made into a master and slave, all of their image caches must have
the same master-slave relationship. This is done automatically. This is also true for
pfMPClipTextures. The pfMPClipTextures that will be the master and slave must be
connected to cliptextures. Only the masters have to be configured, however. When the
other pfMPClipTexture becomes a slave, it configures its cliptexture and makes it and its
image caches slaves as well.
Multipipe Cliptexture API
OpenGL Performer tries to make multipipe cliptexturing as transparent as possible.
Simply call setMaster() on a cliptexture and pass it a pointer to the cliptexture that should
be its master:
•
pfMPClipTexture *slave_mct = pfNewMPClipTexture()
•
pfClipTexture *slave_ct = pfNewClipTexture()
•
pfMPClipTextureClipTexture(slave_mct, slave_ct)
•
pfMPClipTextureMaster(slave_mct, master_mct)
master_mct is a pfMPClipTexture that is already configured.
At this point, slave_mct and master_mct are connected; slave_mct is configured to match
master_mct and shares its image cache resources. The cliptextures and image caches are
also configured and linked. To make pfClipTextures or pfImageCaches masters and
slaves, use the same procedure.
Attaching a pfMPClipTexture to a pfPipe with pfAddMPClipTexture() provides
automatic multipipe support. If a pfMPClipTexture is added to a pipe that is already
connected to another pipe, the function silently creates a new pfMPClipTexture, makes it
a slave of the pfMPClipTexture that is already connected to another pipe, and adds the
slave to the pipe in place of the one passed as an argument to the function.
Multipipe Utilities
Although it is not difficult to set up master and slave cliptextures directly, it is usually not
necessary.The previously described utility routines, pfuAddMPClipTextureToPipes()
and pfuAddMPClipTexturesToPipes() can take multiple pipe arguments. A master pipe
and a list of slave pipes is specified. The routine makes the pfMPClipTexture a master
and attaches it to the master pipe. It then creates slave pfMPClipTextures, attaches them
to the master cliptexture, and attaches a slave cliptexture to each pipe in the slave pipes
007-1680-100
585
15: ClipTextures
list. This routine does extra checking of pipe and cliptexture state, and is guaranteed not
to generate errors, even if the function is applied more than once.
Master/Slave Share Masks
A group of cliptextures grouped by master-slave relationships can do more than share
mem region resources. By default, slave cliptextures also track a number of their master’s
attribute values. This means changing a master’s center, for example, automatically
causes the slaves to change their center locations to match their master’s. The attributes
that a slave can track are divided into groups called share groups. The application can
control which groups are shared by setting a slave’s share mask. Changing the sharing of
a slave only affects that slave’s sharing with its master. Changing the master share mask
has no effect. The share mask is set with the following call:
pfMPClipTextureShareMask(uint mask)
The mask can be set using one or more of the following values:
•
PFMPCLIPTEXTURE_SHARE_CENTER—Slaves track the master’s center.
•
PFMPCLIPTEXTURE_SHARE_DTR—Slaves track DTR: DTR mode, tex load time
(actual or calculated), fade count, and blur margin.
•
PFMPCLIPTEXTURE_SHARE_EDGE—Slaves track texture level parameters,
LODbias invalid border.
•
PFMPCLIPTEXTURE_SHARE_LOD—Slaves track minLOD and maxLOD.
•
PFMPCLIPTEXTURE_SHARE_VIRTUAL—Slaves track lodOffset and num
effective levels.
•
PFMPCLIPTEXTURE_SHARE_DEFAULT—A bit-wise OR of all the masks listed
above.
PFMPCLIPTEXTURE_SHARE_DEFAULT is the default share-mask value, which
provides maximum sharing between master and slave cliptextures. If an application
would like to control one or more slaves independently, it needs to change the slave’s
share mask; then start setting the slaves parameters directly as needed.
Texture Memory and Hardware Support Checking
At the first application or formatting of a cliptexture, OpenGL Performer compares the
expected size of the cliptexture texel data in texture memory against the systems texture
586
007-1680-100
Manipulating Cliptextures
memory size. If it looks like the cliptexture will not fit into texture memory, it shrinks the
clip size by two and tries again. It will keep shrinking the clip size until either the
cliptexture will fit or the clip size is zero. The system takes into account texture memory
banking and paging to come up with a more accurate estimate.
Note that the resizing mechanism does not take into account other textures or
cliptextures in use by the application. You should adjust your application so that
OpenGL Performer does not have to auto-shrink the cliptexture. See “Estimating
Cliptexture Memory Usage” on page 605 for calculating cliptexture system memory and
texture memory usage.
During the checking phase, OpenGL Performer also checks to see if cliptextures are
supported in hardware. If cliptexturing is not supported, one of two emulation modes
will be selected: PFCTEMODE_FRAGPROG or PFCTEMODE_BASIC.
PFCTEMODE_FRAGPROG uses ARB fragment programs to blend imagery stored in
multiple texture units and is automatically selected on systems that support the ARB
fragment program extension. The BASIC cliptexture emulation mode is selected on all
other platforms and only requires OpenGL 1.0 functionality. Both emulation modes work
by transparently assigning cliptextured geometry (pfGeoSets) to dedicated pfDrawBins
that are managed internally within the library. Pre- and post-draw callbacks for such
pfDrawBins are used to override and restore graphics state and to render cliptextured
geometry using the cliptexture emulation state.
Manipulating Cliptextures
Once cliptextures have been configured and connected into the application, they can be
manipulated by the application in the APP process. Applying and centering cliptextures
happens each frame, and is usually an automatic process, set up during post-load
configuration. Other parameters that can be adjusted include load control parameters,
min and max LOD levels, and virtual cliptexture control. Some of these parameters may
only need to be set once in the application, others, like the parameter setting for virtual
cliptextures, need to happen multiple times per frame.
Cliptexture Load Control
The virtualization of pfTextures into pfClipTextures, allowing very large texture maps,
comes at a price. As the clipcenter moves, cliptextures have to download data from disk
to system memory, and from system memory to texture memory. Because of these
007-1680-100
587
15: ClipTextures
download requirements, cliptextures are sensitive to available system bandwidth.
Without some sort of download load control, a fast moving center would cause a
cliptextures to “freeze”, waiting for the system to catch up with its updates.
While mem region updates happen asynchronously, tex region updates must happen in
the DRAW process, competing with geometry rendering and individual texture loading.
Real time applications require that cliptextures, like other OpenGL Performer features,
must be controlled in a way such that an upper bound can be set on their use of resources.
OpenGL Performer’s cliptexture load control, called Dynamic Texture Resolution (DTR),
provides this functionality.
Dynamic Texture Resolution
Dynamic Texture Resolution (DTR) is similar to Dynamic Visual Resolution (DVR): the
bandwidth requirements are adjusted to meet system limitations by lowering the
resolution of the texture data displayed by the cliptexture.
DTR controls bandwidth by analyzing the cliptexture in the CULL process. It checks each
cliptexture level, ensuring that the mem region contains updated tiles corresponding to
the tex region, and that there is enough time to update the tex region within the
download time limit.
This checking goes from level to level, from coarser levels to finer ones. When a level is
found that cannot be displayed, DTR adjusts the cliptexture parameters so that no levels
above the finest complete level are displayed. At that point, DTR stops checking levels
until the next frame. In order not to waste CULL processing time on levels that are not
visible, DTR will not try to sharpen more than one level beyond the current minLOD and
virtualLODoffset values. It will go one level beyond these values so that it can react
quickly if the values change.
In this way the cliptexture updating will always keep up with the movement of the
clipcenter, and will never display invalid data. When the center moves too quickly, DTR
will “blur down” to coarser complete levels, then “sharpen up” to finer levels when the
center slows down and the system can catch up. In this way DTR can trade visual quality
against updating bandwidth. The visual result is that the faster a viewer goes, the less
time there is to download texture and the blurrier the texture data gets.
The nature of cliptexturing makes load control work. When the clipcenter moves, this
change is reflected at every clipped level of the cliptexture. But because each texel in a
level covers four times the geometry of the texel in the next finer level, the clipcenter only
588
007-1680-100
Manipulating Cliptextures
moves half the distance each time you go down a level. This translates into less
demanding texture download requirement.
DTR has other features, such as read queue sorting, which prioritize the order in which
read requests are done to improve mem region update performance. The rate at which
levels are blurred and sharpened can also be controlled to minimize visual artifacts.
Load Control API
DTR controls three aspects of load control, which can be turned on and off
independently: tex region updating (from the mem region in system memory), mem
region updating (loading from disks), and read queue sorting (reducing the latency of
read requests for downloads from disk to the mem region).
pfMPClipTextureDTRMode(DTRMode)
DTRMode is a bitmask; if a bit is set, that DTR feature is enabled. It has the following bits
defined:
•
PF_DTR_MEMLOAD - Enable mem region load control from disk.
•
PF_DTR_TEXLOAD - Enable tex region load control; DRAW download time.
•
PF_DTR_READSORT - Enable priority sorting of the read queue.
All three bits are enabled by default, which means that DTR has all modes enabled.
Besides the bitmask to control what parts of DTR are enabled, there are parameters to
available to adjust load control performance. The DTR parameters and how they affect
DTR functionality are discussed the following subsections.
Download Time
The memload component of DTR is relatively simple; it computes whether all the tiles in
a level’s mem region that cover the tex region are valid. If any are not, the tex region
cannot be updated and DTR invalidates that level. If the texload component of DTR is
enabled, DTR must also compute the time it takes to download from the mem region to
the tex region. The application provides the load control with a total download time in
milliseconds:
pfMPClipTextureTexLoadTime(float _msec)
This is the total time DTR has available to update the cliptexture’s texture memory each
frame. As DTR analyzes each cliptexture level that needs updating, it computes all the
007-1680-100
589
15: ClipTextures
regions in the level’s texture memory that need updating.If a level can be updated, DTR
determines whether there is enough download time left to update the level. If there is,
DTR marks that level valid, subtracts the time needed to download that level from the
total, and starts analyzing the next finest level in the cliptexture.
Cost Tables
OpenGL Performer contains texture download cost tables, which DTR uses to estimate the
time it will take to carry out those texture subloads. These tables are a 2D array of floating
point values, indexed by width and height of the texture rectangle being subloaded. The
cost tables themselves are indexed by machine type and can be read by the application.
The application can also define its own cost tables and configure the system to use it. The
cost table API is shown below:
pfPerf(int type, void *table)
pfQueryPerf(int type, void **table)
The text field indicates whether the cost table should be the one chosen by the system:
•
PFQPERF_DEFAULT_TEXLOAD_TABLE - The one supplied by the application
•
PFQPERF_USER_TEXLOAD_TABLE - The one currently in use.
•
PFQPERF_CUR_TEXLOAD_TABLE - The default table is the current one unless a
application supplies a cost table, in which case, the application’s cost table takes
precedence.
For more details on cost tables, see the man pages for pfPerf() and pfQueryPerf(). The
cost table structure is named pfTexSubloadCostTable, defined in
/usr/include/Performer/pr.h for IRIX and Linux and in
%PFROOT%\Include\pr.h for Microsoft Windows.
Changing Levels
The DTR load control system is designed to minimize visual artifacts as it adjusts for
different download demands. Instead of abruptly sharpening the texture as new levels
with valid texture data become available, DTR blurs in new levels over a number of
frames, making the process of load control less noticeable. The application can control
the rate at which newly valid levels are displayed. The application sets a fade count, which
controls the number of frames it takes to fade in a new level . Each frame, the cliptexture
will sharpen 1/fadecount of the way from its current (possible fractional) level to the next
level. This process is repeated each frame, resulting in an exponential fade-in function. If
the fade count is 0, then fading is disabled, and DTR will show new levels immediately.
590
007-1680-100
Manipulating Cliptextures
pfMPClipTextureFadeCount(int _frames)
If the clipcenter roaming speed leaves barely enough bandwidth to bring in a new
cliptexture level, a distracting “LOD flicker” between two cliptexture LOD levels can
result. Since DTR must blur immediately if a level becomes invalid, the only way to
prevent flicker is to be conservative when sharpening, building in a hysteresis factor. The
parameter called blur margin helps determine when DTR should sharpen.
The blurmargin parameter also helps cliptextures blur smoothly when DTR cannot
keep up. It is a floating point value, which can be interpreted as a fraction of the
cliptexture’s tex load time. When blurmargin is not zero, DTR will load all the levels it
can within the texload time, but not display all of them. Instead it will only sharpen to
the level that would have been reached if the texload time was scaled by blurmargin.
This leaves a cushion of extra time that can be used up before DTR will be forced to blur
to a coarser level. The default blurmargin value of .5 usually causes the finest level
displayed to be one level coarser then the finest level loaded.The application can adjust
blurmargin with this call:
pfMPClipTextureBlurMargin(float margin)
DTR needs this cushion in order to fade smoothly. A cliptexture can only fade between
two valid levels; if it waits until its current level is invalid, the cliptexture must
immediately jump to the next coarser level or it will show invalid data. This abrupt
blurring is very noticeable. The blur margin allows the DTR system to anticipate when it
will lose a level, and smoothly fade to the next coarser level over a number of frames.
Total Texload Time and Texload Time Fraction
Using the texload time, blur margin, and fade count parameters is sufficient to control a
single cliptexture from a pipe, but the interface is awkward if multiple cliptextures are
applying from the same pipe. Since each pipe has the same amount ofDRAW process
time available per frame, no matter how many cliptextures are applied from it, it would
be more convenient to provide a total amount of download time, then divide it among
the cliptextures using the pipe.
OpenGL Performer provides this interface using the total texload time and texload time
fraction parameters. The application can set the total texture download time available on
a pipe, then assign fractional values for each cliptexture, indicating how the download
time should be divided. The total texload time is a pfPipe call, while the fractional values
are set on pfMPClipTextures:
007-1680-100
591
15: ClipTextures
pfPipeTotalTexLoadTime(float msecs)
pfMPClipTextureTexLoadTimeFrac(float frac)
The fractional values should indicate the relative priority of each pfMPClipTexture on
the pipe. The fractional values do not have to add up to 1; the DTR code will normalize
them against the sum of all the fractional values set on the pipe’s pfMPClipTextures.
The total tex load time on the pipe is scaled by the normalized fractional value on each
cliptexture. The scaled tex load time is then used as the cliptexture’s texture download
time. Explicitly setting the tex load time on a pfMPClipTexture will override the
computed fractional time.
Read Queue Sorting
When the clipcenter moves quickly, the number of read requests for texture data tiles that
move into the clipped levels mem regions can grow much faster than the read function
can service them. If there is not enough bandwidth to display a particular level, its read
requests may become “stale”, becoming obsolete as the location of the requested tile
moves into, then passes out of a level’s mem region.For DTR to be robust, the read queue
must be culled and sorted to remove stale read requests, and move the requests for tiles
closest to the clipcenter to the front of the queue. The cliptexture’s read queue is a sorting
queue, which means that a function can process the elements of the queue
asynchronously. DTR uses the read queue to cull read requests for tiles that are no longer
in their mem region, and to prioritize the other requested tiles as a function of level and
distance from the clipcenter. Sophisticated applications can provide their own sorting
function.
Invalidating Cliptextures
Sometimes an application may want to force a cliptexture to completely reload itself. For
example, The pfuGridifyClipTexture() function modifies the cliptexture’s texel data in
system memory with a system of grid marks to make debugging and analysis easier. It
modifies the read function to add a grid to every tile as it’s loaded into system memory,
then invalidates the cliptexture. For more information on gridify, look at the source code
in the /usr/share/Performer/src/lib/libpfutil/gridify.C file for IRIX and
Linux and in %PFROOT%\Src\lib\libpfutil\gridify.C for Microsoft Windows.
Invalidating a cliptexture forces it to completely reload its texture memory. Invalidating
is only supported for cliptextures, not MPcliptextures. This means that an application
592
007-1680-100
Manipulating Cliptextures
cannot call invalidate from the APP process. Instead, it must call invalidate from the
CULL process, usually in a pre-cull callback. The invalidate call itself is simple:
pfInvalidateClipTexture(pfClipTexture *cliptex)
Invalidation is not needed for normal operation, but it is useful as a way to immediately
update a cliptexture’s texture memory.
Virtual ClipTextures
Note: Emulated cliptextures on systems other than InfiniteReality systems are never
virtual.
Regular cliptextures limit the size of each level but do not restrict the number of levels
you can access. Virtual cliptextures take the virtualization a step further by allowing you
to use only a subset of all the levels for which you have data.
Although InfiniteReality supports cliptextures of virtual size up to 8Mx8M = 2^23x2^23
texels (that is, 24 levels), the hardware is only capable of addressing a region of at most
32Kx32K = 2^15x2^15 texels (that is, 16 levels). By limiting the set of texture MIPMap
levels, the cliptextures can be enlarged. A larger, virtual, cliptexture is defined just like a
normal cliptexture, except that the size of the cliptexture can exceed the 32K X 32K
maximum level size dictated by the hardware.
Virtual cliptextures do use more texture memory and require more callbacks in the CULL
process, but they allow enormous cliptextures that are limited only by the precision of
the texture coordinates. Cliptextures over one million texels on a side have been
demonstrated.
Although virtual cliptextures require dividing the cliptextured geometry into sections for
a given MIPmap levelrange, the division is much coarser and less restrictive than texture
tiling. Cliptextured geometry usually does not need to be clipped to sectional
boundaries, for example, since there is a lot of leeway when there are more MIPmap
levels available than are needed for a given section of geometry.
For a sample application implementing virtual cliptextures, see
/usr/share/Performer/src/pguide/libpf/C/virtcliptex.c for IRIX and
Linux.
007-1680-100
593
15: ClipTextures
Selecting the Levels
The application is responsible for choosing which 16 (or less) levels can be accessed at
any given time by setting two parameters: virtualLODOffset and numEffectiveLevels. Most
applications make numEffectiveLevels the maximum number allowed by the hardware, 16
on InfiniteReality. Smaller values may be chosen in some cases to improve stability.
VirtualLODOffset sets the initial level in the cliptexture where 0 is the finest level.
For example, if numEffectiveLevels = 16 and virtualLODOffset = 0 then the texels the
hardware can access are limited to the 32Kx32K region surrounding the current
clipcenter, measured in finest-level texels (actually somewhat less than this. On IRIX and
Linux, see the file /usr/share/Performer/doc/clipmap/IRClipmapBugs.html
or /usr/share/Performer/doc/clipmap/IRClipmapBugs.txt for details on
cliptexture limitations on InfiniteReality graphics). On Microsoft Windows, see the file
%PFROOT%\Doc\clipmap\IRClipmapBugs.html or
%PFROOT%\Doc\clipmap\IRClipmapBugs.txt. Attempting to access outside this
range results in the value of the nearest texel in the good region; that is, the texels forming
the border of the 32Kx32K area will appear to be “smeared” out to fill the virtual
cliptexture.
Increasing virtualLODOffset from 0 to 1 doubles the size of the accessible region in both S
and T (so that it is 32Kx32K level 1 texels, which are twice as big as level 0 texels) but
makes the finest level inaccessible.
The maximum virtualLODOffset allowable is numVirtualLevels-numEffectiveLevels; when
set to that value, the entire S,T range of the virtual cliptexture is accessible, and the finest
level from which texels are available is the 32Kx32K level.
In general, it is appropriate to choose a large value of virtualLODOffset when the
viewpoint is far away from the scene and more S,T area is visible; smaller values of
virtualLODOffset are appropriate as the eye moves closer to the scene, gaining needed
higher resolution at the expense of range in S,T.
Changing virtualLODOffset and numEffectiveLevels has no effect on the contents of texture
memory nor any effect on the texture coordinates stored in the geosets and passed to the
graphics: the texture coordinates, as well as the clipcenter, are always expressed in the
space of the entire virtual cliptexture rather than the smaller “effective” cliptexture of up
to 16 levels within it. (In contrast, changing the clipcenter requires texture downloading;
thus it is a much more expensive operation and therefore it is not practical to change the
clipcenter more than once per frame, whereas virtualLODOffset and numEffectiveLevels
can be changed multiple times per frame, as we will see in the following subsections.)
594
007-1680-100
Manipulating Cliptextures
How to Set Virtual Cliptexture Parameters
OpenGL Performer supports two different methods for managing virtualLODOffset and
numEffectiveLevels of a cliptexture. The simpler of the two methods allows the parameters
to be set and changed at most once per frame; the more sophisticated method allows
them to be changed multiple times per frame (different values for different parts of the
scene). In addition to virtualLODOffset and numEffectiveLevels described earlier, the
parameters minLOD, maxLOD, LODBiasS and LODBiasT often need to be set in the
same way; so, we will show how to set those as well.
Per-Frame Setting of Virtual Cliptexture Parameters
The easy way to manage the virtual cliptexture parameters is to set the values of the
parameters on the pfMPClipTexture controlling the pfClipTexture:
int LODOffset, numEffectiveLevels;
float minLOD, maxLOD;
float LODBiasS, LODBiasT, LODBiasR;
...
mpcliptex->setVirtualLODOffset(LODOffset);
mpcliptex->setNumEffectiveLevels(numEffectiveLevels);
mpcliptex->setLODRange(minLOD, maxLOD);
mpcliptex->setLODBias(LODBiasS, LODBiasT, LODBiasR);
You make these calls in the APP process, either in the main program loop, a channel APP
func, or a pre- or post-node APP func. The last value you give during the APP in a
particular frame will be used for rendering that frame and all subsequent frames until
you change the value again.
This simple technique is the one that is used by the clipfly program when you
manipulate the LODOffset and EffectiveLevels sliders (when using a naive scene loader
such as the .im loader that does not do its own management of virtualLODOffset and
numEffectiveLevels): clipfly makes these calls in its channel pre-APP function.
This technique is also used by the .spherepatch loader; in this case, the calls are made
in a post-APP function of a node in the scene graph, using parameters that are
intelligently chosen based on the current distance from the eye to the closest point on the
textured geometry and are updated every frame.
Notice that even though the .spherepatch loader manages the virtualLODOffset and
numEffectiveLevels, you can still modify or override its behavior with the clipfly GUI
controls. This is accomplished using a special set of “limit” parameters that are provided
007-1680-100
595
15: ClipTextures
as a convenience and stored on the pfMPClipTexture. The intended use is for
applications such as clipfly to set the limits based on GUI input or other criteria:
mpcliptex->setLODOffsetLimit(lo, hi);
mpcliptex->setEffectiveLevelsLimit(lo, hi);
mpcliptex->setMinLODLimit(lo, hi);
mpcliptex->setMinLODLimit(lo, hi);
mpcliptex->setLODBiasLimit(Slo, Shi, Tlo, Thi, Rlo, Rhi);
Then the callback functions of intelligent loaders such as .spherepatch query the
limits:
mpcliptex->getLODOffsetLimit(&lo, &hi);
mpcliptex->getEffectiveLevelsLimit(&lo, &hi);
mpcliptex->getMinLODLimit(&lo, &hi);
mpcliptex->getMinLODLimit(&lo, &hi);
mpcliptex->setLODBiasLimit(&Slo, &Shi, &Tlo, &Thi, &Rlo, &Rhi);
The loaders use the limits to modify the selection of the final parameters sent to
pfMPClipTexture.
The limits are not enforced by pfMPClipTexture; they are provided merely to facilitate
communication from the application to the function controlling the parameters. That
function is free to ignore or only partially honor the limits if it wishes.
The limits may also be queried frame-accurately from the pfMPClipTexture in the CULL
process, so they can also be used by scene loaders such as the .ct loader that use the
per-tile method described in the next section.
Per-Tile Setting of Virtual Cliptexture Parameters
Many applications require accessing a wider range of the cliptexture’s data than can be
obtained by a single setting of virtualLODOffset and numEffectiveLevels. This can be
accomplished by partitioning the database into “tiles” roughly according to distance
from the eye or from the texture’s clipcenter and setting the parameters for each tile every
frame in the pre-CULL func of the pfGroup or pfGeode representing that tile by calling
pfClipTexture::applyVirtual(), pfTexture::applyMinLOD(),
pfTexture::applyMaxLOD(), and pfTexture::applyLODBias().
596
007-1680-100
Manipulating Cliptextures
Tiling Strategies
Choosing a database tiling strategy requires careful thought and tuning. The most
conceptually straightforward method is to use a static 2D grid-like spatial partitioning.
This method requires tuning the granularity of the partitioning for the particular
database and capabilities of the machine: if a tile is too big and sufficiently close to the
eye, there may be no possible combination of virtualLODOffset and numEffectiveLevels
that allows access to both the necessary spatial range and texture LOD range without
garbage in the distance or excess bluriness in the foreground; but if there are too many
tiles, the overhead of changing the parameters for each tile can become excessive.
In general, assuming the maximum active area is 32Kx32K (as it is on InfiniteReality),
each tile should be small enough so that it covers at most approximately 16K texels at the
finest texture LOD that will be used when rendering it; this is so that when the clipcenter
is close enough to the tile to require accessing that finest texture LOD, the 32Kx32K good
area centered at approximately the clipcenter will be able to cover it with some slop left
over to account for the inexact placement of the good area (see the IR cliptexture bugs
doc). (Finer tiles such as 8Kx8K or even 4Kx4K can be used for improved stability under
extreme magnification; see the IR cliptexture bugs doc).
This rule has two important consequences:
•
If your cliptexture has insets (that is, localized regions in which higher-resolution
data is available) you can make the tiling coarser in the regions where only
low-resolution data is available and finer at the insets.
•
If you use pfLODs to optimize your database, the coarse LODs of the pfLOD can
(and should) be tiled more coarsely than the fine ones.
This is because the coarser LODs are used at far distances, and at those far distances
the Mipmapping hardware will only want to access correspondingly coarse texture
levels anyway, so the 16Kx16K can be measured in terms of the texels of those
coarse texture levels.
A more general tiling strategy that requires less empirical database tuning than the static
tiling method is to make the tiles be concentric rings around the texture’s clipcenter (in
2D) or around the eye point (in 3D), with sizes increasing in approximately powers of 2.
However, since the clipcenter and view position changes, this means the tiles must move
as well, which requires dynamically changing the topology of the scene graph and/or
morphing the geometry so that the tiles always form those concentric rings around the
current focus.
007-1680-100
597
15: ClipTextures
The .ct loader and pfASD’s ClipRings both use this dynamic strategy. The .ct loader
is interesting in that the morphing is done for the sole purpose of forming these
concentric tiles for virtual-cliptexturing an otherwise trivial scene. It looks like simply a
square textured by the cliptexture, but if you turn on scribed mode in perfly or clipfly,
you can see the morphing rings that make up the square.
Doing Per-tile Updates
To do per-tile updates, use the following procedure:
1.
On each tile (typically a pfGroup or pfGeode) put a pre-node CULL func:
tile->setTravFuncs(PFTRAV_CULL, tilePreCull, NULL);
2. Make sure the effect of the tile’s pre-CULL func happens in the DRAW before the
contents of the tile are rendered, and that the tile’s contents do not co-mingle with
other tiles (this is not guaranteed by default, for the benefit of CULL whose sole
purpose is to return a CULL result without losing the advantages of uncontained
CULL sorting):
tile->setTravMode(PFTRAV_CULL, PFTRAV_CULL_SORT,
PFN_CULL_SORT_CONTAINED);
3. In the pre-node CULL func for the tile, set the parameters:
static int tilePreCull(pfTraverser *trav, void *)
{
int virtualLODOffset, numEffectiveLevels;
float minLOD, maxLOD;
float biasS, biasT, biasR;
//Choose intelligent values for parameters.
cliptex->applyVirtualParams(virtualLODOffset,
numEffectiveLevels);
cliptex->applyMinLOD(minLOD);
cliptex->applyMaxLOD(maxLOD);
cliptex->applyLODBias(biasS,biasT,biasR);
}
The values given to the apply functions are not stored in the pfClipTexture or retained
from frame to frame; when you call these functions, they override the corresponding
values stored in the cliptexture.
598
007-1680-100
Manipulating Cliptextures
It is not necessary to call all four of the apply...() functions; only use the ones you care
about (for example, most applications would not care about LODBias). However, if you
ever call a given one of these functions, applyMinLOD(), for example, on a particular
cliptexture for any tile, then you must call applyMinLOD() for every tile on that
cliptexture during that frame and forever after; if you omit it, the tile will not necessarily
get the value stored on the pfMPClipTexture or pfClipTexture; rather, it will get whatever
value happened to be most recently set when rendering that tile in the DRAW (which
may be nondeterministic due to CULL sorting of the scene graph).
How to Choose Virtual Cliptexture Parameters
The libpfutil library provides the function pfuCalcVirtualClipTexParams(),
which can be very useful in selecting the virtual cliptexture parameters, regardless of
whether you are updating per-frame or per-tile.
Essentially, you give to pfuCalcVirtualClipTexParams() every piece of information you
know about the cliptexture:
•
the tile in question
•
the limits specified elsewhere, for example, by the clipfly GUI
pfuCalcSizeFinestMipLOD() returns the lower bounds on minLODPixTex, which is one
of the input parameters to pfuCalcVirtualClipTexParams().
The function returns optimal values for virtualLODOffset, numEffectiveLevels, minLOD,
maxLOD. You can do the following with them:
•
Set on the pfMPClipTexture in the APP process if your application is using the
per-frame method.
•
Apply to the pfClipTexture per-tile in the CULL process if using the per-tile method.
For more details, you may also want to read the commented source code to understand
its constraints and heuristics, and how to modify pfuCalcVirtualClipTexParams to
implement your own algorithm if it does not exactly suit your needs.
Custom Read Functions
Sometimes the read function supplied by OpenGL Performer to download texture data
from disk to mem region is not good enough. The application may need to do additional
operations at read time, such as uncompression, or may need a more sophisticated read
007-1680-100
599
15: ClipTextures
function, such as an interruptible one for reading large tiles from slow storage devices. A
read function may need to signal an applications secondary caching system; for example,
reading from tape storage to disk.
OpenGL Performer provides support for application supplied custom read functions.
The read function is supplied at configuration time, and there is API in both the
configuration utilities and the cliptexture and image cache configuration files for
supplying a read function.
A read function is called by the image caches read queue. The read queue expects a read
function with the following function signature:
int ExampleReadFunction(pfImageTile *it, int ntexels)
The image tile pointer provides information about the read request, such as the disk to
read from, the dimensions and format of the texel data, and the destination tile in system
memory to write to. The ntexels argument is an integer indicating the number of texels to
read from disk. The read function returns another integer indicating the number of texels
actually read. Two example read functions, ReadNormal() and ReadDirect(), are
supplied in /usr/share/Performer/src/lib/libpfdu/pfdLoadImage.c for
IRIX and Linux and in %PFROOT%\Src\lib\libpfdu\pfdLoadImage.c for
Microsoft Windows. These functions are C versions of the C++ functions that OpenGL
Performer uses to read texture data. In OpenGL Performer, the ReadDirect() function is
called by the read queue; it tries to use direct I/O to get the highest possible disk read
performance. If the read direct call fails, it calls ReadNormal(), which uses normal
fopen()-style read.
When providing a read function at configuration time, You supply the function name,
and optionally the name of a DSO library containing the function. If no dynamic shared
library is supplied, the read function is searched for in the application’s executable.
To set custom read functions using the configuration utilities, simply fill in the readFunc
field in the pfuImgCacheConfig or pfuClipTexConfig structure (the first structure has
priority over the second if both are set). The field should contain a pointer to the custom
read function. Be sure the function has the proper signature.
When supplying custom read functions in the configuration files, you simply provide an
entry in one of two formats:
read_func ReadFunctionName
read_func DSOlibraryName ReadFunctionName
600
007-1680-100
Using Cliptextures
For hints on when and how to use custom read functions, see the customizing read
functions in “Custom Read Functions” on page 612.
Using Cliptextures
This section provides guidelines for using cliptextures, describing common cliptexture
application techniques, ways to solve problems, and some hints and tips to make using
cliptextures easier.
Cliptexture Insets
Cliptexture load control makes it possible to create cliptextures with incompletely filled
levels. A cliptexture, being much larger than an ordinary texture, may not be used in a
homogeneous way. Some areas of the cliptexture may be viewed in detail, others only at
a distance. A good example of this usage pattern is flight simulation. The terrain around
an airport will be seen from low altitude, terrain far from population centers may never
be seen below 40,000 feet. It is also possible that high resolution data is simply not
available for the entire cliptexture. Both of these cases make it valuable to create
cliptextures with incompletely populated levels.
Regions of filled in data are called insets. Insets can be any shape, and do not need to
match tile boundaries (although this requires filling the rest of the tile with super
sampled data). For an inset to work properly, all of the levels from the pyramid up to the
finest level desired, must be available within the inset boundaries.
007-1680-100
601
15: ClipTextures
Insets
Figure 15-12
Cliptexture Insets
Insets are supported in cliptextures as a natural consequence of load control. As the
clipped region moves from a region that has texel data to one that does not, DTR will blur
the texture down to the highest level that can completely fill the clipped region.
Adding Insets to Cliptextured Data
In large cliptextures, it may not be practical or even desirable to completely fill each level
with texel data. Cliptexture’s load control, DTR, automatically adjusts the finest visible
level based on what texels are available. If finer levels are not available, DTR
automatically “blurs down” to the highest complete level in the clip region.
Applications may use insets if there are only limited areas where the viewer is close to
the terrain. An example application would be a commercial flight simulator, where the
inset high-resolution data would be around the airports where the aircraft takes off and
lands. The terrain over which the aircraft cruises can be lower resolution.
Insets and DTR
To create insets properly, you have to understand how DTR load control works. At the
beginning of each frame, DTR examines a level’s mem region to see if the tiles covering
the tex region are all loaded. If the tiles are all available, DTR will make that level visible.
602
007-1680-100
Using Cliptextures
DTR