007 1680 080

User Manual: 007-1680-080

Open the PDF directly: View PDF .
Page Count: 842

Download
Open PDF In Browser	View PDF

OpenGL Performer™
Programmer’s Guide

007-1680-080

CONTRIBUTORS
Written by George Eckel and Ken Jones
Illustrated by Chrystie Danzer and Chris Wengelski
Production by Glen Traefald
Engineering contributions by Angus Dorbie, Paolo Farinelli, Tom Flynn, Yair Kurzion, Radomir Mech, Alexandre Naaman, Marcin
Romaszewicz, Allan Schaffer, and Jenny Zhao

COPYRIGHT
© 1994, 2000–2002 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 electronic (software) version of this document was developed at private expense; if acquired under an agreement with the USA government
or any contractor thereto, it is acquired as "commercial computer software" subject to the provisions of its applicable license agreement, as
specified in (a) 48 CFR 12.212 of the FAR; or, if acquired for Department of Defense units, (b) 48 CFR 227-7202 of the DoD FAR Supplement; or
sections succeeding thereto. Contractor/manufacturer is Silicon Graphics, Inc., 1600 Amphitheatre Pkwy 2E, Mountain View, CA 94043-1351.

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, OpenGL Multipipe, OpenGL Performer,
OpenGL Shader, Performance Co-Pilot, REACT, RealityEngine, RealityEngine2, Showcase, and VPro are trademarks of Silicon Graphics, Inc.
AutoCAD is a registered trademark of Autodesk, Inc. 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. 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. RPC is a trademark of ArchVision. VTune is
a trademark of Intel Corporation. Wavefront is a trademark of Alias|Wavefront, a division of Silicon Graphics Limited. 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.

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 3.0, which has the following
features:

007-1680-080

•

New hardware support
- SGI Scalable Graphics Compositor
- InfiniteReality4

•

New OS platforms
- Windows NT 4.0
- Windows 2000
- Windows XP

•

Functional features
- Image-Based rendering using simplified 3D proxy geometry
- Other image-based rendering enhancements
- Occlusion culling using cull programs
- Cull helper processes (cull sidekick)
- New library libpfv, providing a feature-rich viewer

•

Integration with other graphics toolkits
- OpenGL Multipipe SDK
- Direct integration with OpenGL Shader 3.0
- OpenGL Volumizer

•

OpenFlight 15.7 support

iii

Record of Revision

Version

Description

020

1994
Original publication.

007-1680-080

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.

Contents

Figures .

. xxvii

Tables .

Examples

. xxxv

.
.
.
.
.
.
.
.
.

. xxxix
. xxxix
. . xl
. . xl
. . xl
. . xlii
. . xlii
. xliii
. xliii

.
.
.
.
.
.
.
.
.
.
.
.
.
.

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.

007-1680-080

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

.
.
.
.
.
.
.
.
.
.
.
.
.
.

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

.
.
.
.
.
.
.
.

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

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

007-1680-080

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.

007-1680-080

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

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

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.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

100
101
101
102
102
103
111
112
115
117
120
120
120
121
122
124
124
125
125
125
126

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

007-1680-080

Contents

007-1680-080

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 . . .
Patchy Fog and Layered Fog .
Creating Layered Fog .
Creating Patchy Fog . .
Initializing a pfVolFog .
Updating the View . .
Drawing a Scene with Fog

.
.
.
.
.
.
.
.
.

.181
.181
.182
.186
.187
.187
.188
.191
.191

.
.
.
.
.
.
.
.
.

Contents

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. . . . . . . . . . . . . . . .
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 . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

191
192
193
196
198
199
201
202
202
203
204
204
205
206
208
209
215
219
220

.
.
.
.
.
.
.
.
.
.
.
.
.
.

221
221
222
223
225
228
229
229
230
230
240
240
241
241

007-1680-080

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 . . . . . .

007-1680-080

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.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

xiii

Contents

WRL Format . . . . . . . .
Database Operators with Pseudo Loaders .

.
.

287
288

Geometry . . . . .
Geometry Sets . . . .
Primitive Types . .
pfGeoSet Draw Mode .
Primitive Connectivity
Attributes . . . .
Attribute Bindings . .
Indexed Arrays . .
pfGeoSet Operations .
3D Text . . . . . .
pfFont . . . . .
pfString . . . . .

.
.
.
.
.
.
.
.
.
.
.
.

291
291
293
294
296
299
301
302
304
305
305
307

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 . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.

311
311
313
318
318
319
335
335
337
338
339
339

Shader . . . . . . . . . .
Setting Up Shaded pfGeoSets . . . .
Textures and Texture Coordinates . .
OpenGL Shader and OpenGL Performer

.
.
.
.

347
347
348
349

10.

xiv

.
.
.
.
.
.
.
.
.
.
.
.

007-1680-080

Contents

11.

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 . . . . . . . . . .
Using the pfCompositor Class . . . . . . . . . .
Load Balancing . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.

.351
.351
.351
.352
.359
.364
.365
.365
.367
.370

12.

ClipTextures . . . . . . . . . . . . .
Overview . . . . . . . . . . . . . .
Cliptexture Levels . . . . . . . . . .
Cliptexture Assumptions . . . . . . . .
Image Cache . . . . . . . . . . . .
Toroidal Loading . . . . . . . . . . .
Updating the Clipcenter . . . . . . . . .
Virtual Cliptextures . . . . . . . . . .
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 . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.373
.374
.375
.376
.377
.380
.382
.382
.384
.384
.385
.386
.387
.388
.389
.389
.391
.391
.392
.392
.393
.393

007-1680-080

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

Contents

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
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 . . . . . . . .
13.

xvi

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

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 . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

393
393
397
399
415
415
417
420
422
423
423
428
429
435
437
437
441
445
446
447
448
448

.
.
.
.
.
.
.
.
.
.
.

453
453
454
457
459
461
462
463
464
464
465

007-1680-080

Contents

14.

pfPipeWindows and pfPipeVideoChannels . . . .
Using pfPipeWindows . . . . . . . . . .
Creating, Configuring and Opening pfPipeWindow .
pfPipeWindows in Action . . . . . . . .
Controlling Video Displays . . . . . . . . .
Creating a pfPipeVideoChannel . . . . . .
Multiple pfPipeVideoChannels in a pfPipeWindow .
Configuring a pfPipeVideoChannel . . . . .
Use pfPipeVideoChannels to Control Frame Rate .

.
.
.
.
.
.
.
.
.

.469
.469
.469
.480
.482
.483
.484
.484
.485

15.

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 . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.487
.487
.488
.488
.489
.490
.493
.494
.495
.495
.496
.497
.498
.498
.499
.502
.502
.504

16.

Dynamic Data . . . . . .
pfFlux . . . . . . . .
Creating and Deleting a pfFlux
Initializing the Buffers . .
pfFlux Buffers . . . . .

.
.
.
.
.

.507
.507
.508
.509
.510

007-1680-080

.
.
.
.
.

xvii

Contents

17.

xviii

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 . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

513
515
518
519
521
522
522
523
530
530
530
531
531
532

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 . .
Data Structures . . . . .
Triangle Data Structure .
Attribute Data Array . .
Vertex Data Structure . .
Default Evaluation Function

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

535
535
537
537
538
540
541
542
543
543
544
546
547
549
554
555
556

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

007-1680-080

Contents

18.

007-1680-080

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 . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.558
.558
.559
.560
.560
.561
.561
.562
.562
.564
.566
.566
.567
.568

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
Preprocessing Light Points . . . .
Stage Configuration Callbacks . .
How the Light Point Process Works

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.571
.571
.572
.573
.573
.574
.574
.576
.577
.578
.579
.579
.582
.584
.585
.585
.586

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

xix

Contents

19.

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 . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

587
588
590
592
593
595
597
599
600
600
601
601
602
602
602
603

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 . .
Creating and Working with Line Segments .
Intersecting with Volumes . . . .
Intersecting with Planes and Triangles .
Intersecting with pfGeoSets . . . .
General Math Routine Example Program .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

611
611
613
617
619
620
620
622
623
624
624
624
626
627
628
628
630

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

007-1680-080

Contents

20.

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 . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.635
.636
.636
.637
.640
.641
.643
.644
.644
.645
.646
.646
.647
.650
.651
.653
.653
.653
.654

21.

Performance Tuning and Debugging . . . . . . .
Performance-Tuning Overview . . . . . . . . .
How OpenGL Performer Helps Performance . . . . .
Draw Stage and Graphics Pipeline Optimizations . .
Cull and Intersection Optimizations . . . . . .
Application Optimizations . . . . . . . . .
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 . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.655
.655
.656
.656
.658
.659
.660
.660
.663
.667
.672
.672
.673
.674

007-1680-080

.
.
.
.
.
.
.
.
.
.
.
.
.
.

xxi

Contents

xxii

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 . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

675
675
675
676
676
677
678
678
679
680
680
681
681
682
682
683
684
684
688
695

22.

Building a Visual Simulation Application Using libpfv. .
Overview . . . . . . . . . . . . . . .
The Simplest pfvViewer Program . . . . . . . .
Adding Interaction to a pfvViewer Program . . . . .
Reading XML Configuration Files . . . . . . . .
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

.
.
.
.
.
.
.
.
.
.

699
699
700
701
702
706
709
711
712
716

23.

Programming with C++ .
Overview . . . . .

.
.

719
719

.
.

007-1680-080

Contents

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 . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.720
.720
.720
.721
.721
.721
.721
.725
.727
.727
.727
.728
.729
.729
.730
.731
.732
.732
.733
.734
.735
.736
.736
.736

Glossary

.737

.759

Index

007-1680-080

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

xxiii

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

007-1680-080

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
.127
.132
.137
.139
.149
.153
.162
.172
.174
.183
.186

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

xxvii

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 8-1
Figure 8-2
Figure 8-3
Figure 8-4
Figure 9-1
Figure 9-2
Figure 11-1
Figure 11-2
Figure 11-3
Figure 11-4
Figure 11-5
Figure 11-6
Figure 12-1
Figure 12-2
Figure 12-3
Figure 12-4
Figure 12-5

xxviii

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 . . .
Primitives and Connectivity . . . . . . . .
pfGeoSet Structure . . . . . . . . . . .
Indexing Arrays . . . . . . . . . . . .
Deciding Whether to Index Attributes . . . . .
pfGeoState Structure . . . . . . . . . .
Generating the Color of a Multi-textured Pixel . . .
pfPipes Creating pfHyperpipes . . . . . . .
Multiple Hyperpipes . . . . . . . . . .
Default Hyperpipe Mapping to Graphic Pipes . . .
Attaching Objects to the Master pfPipe . . . . .
Hardware Composition Schemes . . . . . . .
Decomposition Modes for pfCompositor. . . . .
Cliptexture Components . . . . . . . . .
Image Cache Components . . . . . . . . .
Mem Region Update . . . . . . . . . .
Tex Region Update . . . . . . . . . . .
Cliptexture Cache Hierarchy . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

210
213
245
250
251
253
255
258
261
262
266
269
271
277
282
284
298
300
303
304
344
345
353
354
355
360
366
367
374
375
378
379
380

007-1680-080

Figures

Figure 12-6
Figure 12-7
Figure 12-8
Figure 12-9
Figure 12-10
Figure 12-11
Figure 12-12
Figure 12-13
Figure 12-14
Figure 14-1
Figure 15-1
Figure 15-2
Figure 16-1
Figure 16-2
Figure 16-3
Figure 16-4
Figure 17-1
Figure 17-2
Figure 17-3
Figure 17-4
Figure 17-5
Figure 17-6
Figure 17-7
Figure 17-8
Figure 17-9
Figure 17-10
Figure 17-11
Figure 17-12
Figure 17-13
Figure 17-14
Figure 17-15
Figure 17-16
Figure 17-17

007-1680-080

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. . . . . . . . . . . .
Example Setup for Geometry Alignment . . . . . . .
Aligning Light Points Above a pfASD Surface . . . . . .

.381
.382
.383
.416
.419
.420
.438
.440
.446
.482
.487
.501
.508
.511
.517
.519
.539
.540
.541
.542
.545
.545
.546
.547
.548
.551
.551
.552
.553
.556
.562
.563
.564

xxix

Figures

Figure 17-18
Figure 18-1
Figure 18-2
Figure 18-3
Figure 18-4
Figure 18-5
Figure 20-1
Figure 20-2
Figure 20-3
Figure 21-1
Figure 21-2
Figure 21-3
Figure 21-4
Figure 21-5

xxx

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 . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

567
572
575
576
583
591
636
638
642
683
688
692
693
694

007-1680-080

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

007-1680-080

Routines that Modify libpr Object Reference Counts
Attributes in the Share Mask of a Channel Group . .

.
.

. 12
. 41

OpenGL Performer Node Types . . . . . . . . . .
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 . . . . . . .

. 51
. 54
. 63
. 63
. 67
. 70
. 71
. 72
. 73
. 75
. 79
. 86
.105
.106
.107
.109
.112
.121
.123
.123
.124
.131
.145
.156
.168

xxxi

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 8-1
Table 8-2
Table 8-3
Table 8-4
Table 8-5
Table 8-6
Table 9-1
Table 9-2
Table 9-3
Table 9-4
Table 9-5
Table 9-6
Table 9-7

xxxii

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 . . . . . .
pfGeoSet Routines . . . . . . . . . . .
Geometry Primitives . . . . . . . . . .
pfGeoSet PACKED_ATTR Formats . . . . . .
Attribute Bindings . . . . . . . . . . .
pfFont Routines . . . . . . . . . . . .
pfString Routines . . . . . . . . . . .
pfGeoState Mode Tokens . . . . . . . . .
pfTransparency Tokens
. . . . . . . . .
pfGeoState Value Tokens . . . . . . . . .
Enable and Disable Tokens
. . . . . . . .
Rendering Attribute Tokens . . . . . . . .
Texture Image Sources . . . . . . . . . .
Texture Load Modes . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

184
184
188
189
190
200
201
215
221
223
225
226
239
242
259
273
275
278
279
289
292
293
296
301
305
308
313
314
318
318
319
322
325

007-1680-080

Tables

Table 9-8
Table 9-9
Table 9-10
Table 9-11
Table 9-12
Table 9-13
Table 11-1
Table 11-2
Table 11-3
Table 12-1
Table 12-2
Table 12-3
Table 12-4
Table 12-5
Table 12-6
Table 13-1
Table 13-2
Table 13-3
Table 13-4
Table 14-1
Table 14-2
Table 15-1
Table 15-2
Table 15-3
Table 15-4
Table 15-5
Table 15-6
Table 15-7
Table 16-1
Table 17-1
Table 18-1
Table 19-1
Table 19-2

007-1680-080

Texture Generation Modes . . . . . . . . . .
pfFog Tokens . . . . . . . . . . . . . .
pfHlightMode() Tokens . . . . . . . . . . .
Matrix Manipulation Routines . . . . . . . . .
pfSprite Rotation Modes . . . . . . . . . . .
pfGeoState Routines . . . . . . . . . . . .
pfPipeWindow Functions That Do Not Propagate
. . .
Functions for Managing Compositors . . . . . . .
Load Balancing Functions . . . . . . . . . . .
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 . . . . . . . . . . .
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 . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.329
.333
.334
.335
.336
.342
.361
.368
.372
.390
.402
.406
.408
.412
.413
.455
.458
.460
.462
.472
.477
.490
.493
.495
.496
.503
.503
.504
.522
.549
.588
.612
.613

xxxiii

Tables

Table 19-3
Table 19-4
Table 19-5
Table 19-6
Table 19-7
Table 19-8
Table 19-9
Table 19-10
Table 19-11
Table 19-12
Table 23-1
Table 23-2
Table 23-3
Table 23-4
Table 23-5
Table 23-6

xxxiv

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 .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

618
619
622
623
623
624
625
625
629
630
720
721
722
723
724
730

007-1680-080

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

007-1680-080

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
.118
.135
.143

xxxv

Examples

Example 5-3
Example 6-1
Example 6-2
Example 6-3
Example 8-1
Example 8-2
Example 9-1
Example 9-2
Example 9-3
Example 9-4
Example 11-1
Example 11-2
Example 11-3
Example 11-4
Example 11-5
Example 12-1
Example 13-1
Example 13-2
Example 13-3
Example 13-4
Example 14-1
Example 14-2
Example 14-3
Example 14-4
Example 14-5
Example 16-1
Example 16-2
Example 17-1
Example 18-1
Example 18-2
Example 18-3
Example 18-4

xxxvi

Default Stress Function . . . . . . . . . . . . .
How to Configure a pfEarthSky . . . . . . . . . .
Fog initialization Using pfVolFogAddPoint() . . . . . .
Specifying Patchy Fog Boundaries Using pfVolFogAddNode() .
Loading Characters into a pfFont . . . . . . . . . .
Setting Up and Drawing a pfString . . . . . . . . .
Using pfDecal() to a Draw Road with Stripes . . . . . .
Pushing and Popping Graphics State . . . . . . . . .
Using pfOverride() . . . . . . . . . . . . . .
Inheriting State . . . . . . . . . . . . . . .
Configuring a System with Three Hyperpipe Groups . . . .
Mapping Hyperpipes to Graphic Pipes . . . . . . . .
More Complete Example: Mapping Hyperpipes to Graphic Pipes
Set FBConfigAttrs for Each pfPipeWindow . . . . . . .
Search the pfPipeWindow List of the pfPipe. . . . . . .
Estimating System Memory Requirements . . . . . . .
Opening a pfWindow . . . . . . . . . . . . .
Using the Default Overlay Window . . . . . . . . .
Creating a Custom Overlay Window . . . . . . . . .
pfWindows and X Input . . . . . . . . . . . .
Creating a pfPipeWindow . . . . . . . . . . . .
pfPipeWindow With Alternate Configuration Windows for
Statistics . . . . . . . . . . . . . . . . .
Custom Initialization of pfPipeWindow State . . . . . .
Configuration of a pfPipeWindow Framebuffer. . . . . .
Opening and Closing a pfPipeWindow . . . . . . . .
Fluxed pfGeoSet . . . . . . . . . . . . . . .
Connecting Engines and Fluxes . . . . . . . . . .
Aligning Light Points Above a pfASD Surface . . . . . .
Raster Callback Skeleton . . . . . . . . . . . .
Preprocessing a Display List - Light Point Process code . . .
Setting pfCalligraphic Parameters. . . . . . . . . .
Calligraphic Lights . . . . . . . . . . . . . .

154
182
187
187
306
307
317
338
339
341
355
356
356
362
363
443
454
465
466
467
470
474
476
479
480
518
532
564
581
586
599
603

007-1680-080

Examples

Example 19-1
Example 19-2
Example 19-3
Example 19-4
Example 19-5
Example 23-1
Example 23-2
Example 23-3
Example 23-4
Example 23-5

007-1680-080

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 . . . . . .

.
.
.
.
.
.
.
.
.
.

.616
.618
.626
.627
.630
.726
.726
.730
.731
.734

xxxvii

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
InfiniteReality, InfiniteReality2, Silicon Graphics Octane, Silicon Graphics O2, and VPro
graphics subsystems. OpenGL Performer teamed with InfiniteReality or Octane 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

007-1680-080

xxxix

About This Guide

graphics platforms; you do 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:
•

Chapter 1, “OpenGL Performer Programming Interface,” describes the
fundamental ideas behind the OpenGL Performer programming interface.

007-1680-080

About This Guide

007-1680-080

•

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, “Graphics State,” describes the graphics state, which contains all of the
fields that together define the appearance of geometry.

•

Chapter 10, “Shader,” describes the shader, a mechanism that allows complex
rendering equations to be applied to OpenGL Performer objects.

•

Chapter 11, “Using Scalable Graphics Hardware,” describes how to use OpenGL
Performer in conjunction with an SGI Video Digital Multiplexer (DPLEX) and an
SGI Scalable Graphics Compositor.

•

Chapter 12, “ClipTextures,” describes how to work with large, high-resolution
textures.

•

Chapter 13, “Windows,” describes how to create, configure, manipulate, and
communicate with a window in OpenGL Performer.

•

Chapter 14, “pfPipeWindows and pfPipeVideoChannels,” describes the unified
window and video channel control and management provided by pfPipeWindows
and pfPipeVideoChannels.

•

Chapter 15, “Managing Nongraphic System Tasks,” describes clocks, memory
allocation, synchronous I/O, error handling and notification, and search paths.

•

Chapter 16, “Dynamic Data,” describes how to connect pfFlux, pfFCS, and
pfEngine nodes, which together can be used for animating geometries.

xli

About This Guide

•

Chapter 17, “Active Surface Definition,” describes the Active Surface Definition
(ASD): a library that handles real-time surface meshing and morphing.

•

Chapter 18, “Light Points,” describes the calligraphic lights, which are intensely
bright lights.

•

Chapter 19, “Math Routines,” details the comprehensive math support provided as
part of OpenGL Performer.

•

Chapter 20, “Statistics,” discusses the various kinds of statistics you can collect and
display about the performance of your application.

•

Chapter 21, “Performance Tuning and Debugging,” explains how to use
performance measurement and debugging tools and provides hints for getting
maximum performance.

•

Chapter 22, “Building a Visual Simulation Application Using libpfv” describes a
modular approach to building an application using a graphical viewer.

•

Chapter 23, “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.

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.

xlii

007-1680-080

About This Guide

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://techpubs.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 info-performer-request@sgi.com. Much like the comp.sys.sgi.*
newsgroups on the Internet, it is not an official support channel but is monitored by
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.)

007-1680-080

xliii

About This Guide

You can contact us in any of the following ways:
•

Send e-mail to the following address:
techpubs@sgi.com

•

Use the Feedback option on the Technical Publications Library World Wide Web
page:
http://techpubs.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

•

Send a fax to the attention of Technical Publications:
+1 650 932 0801

We value your comments and will respond to them promptly.

xliv

007-1680-080

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

007-1680-080

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
#include
.....
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.

007-1680-080

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:

007-1680-080

•

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

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.

007-1680-080

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);

007-1680-080

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.

007-1680-080

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.

007-1680-080

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.

007-1680-080

Base Classes

pfObject
pfLight
pfPipe

pfMaterial
pfNode

pfGeoSet

pfChannel
pfFrustum

Some classes
found in libpf

Some classes
found in libpr

Figure 1-1

007-1680-080

Partial Inheritance Graph of OpenGL Performer Data Types

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.

007-1680-080

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.

007-1680-080

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().

007-1680-080

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 */

007-1680-080

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.
•

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().

007-1680-080

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:
•

007-1680-080

PFPRINT_VB_OFF—no printing

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.

007-1680-080

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().

007-1680-080

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 22, “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.

007-1680-080

2: Setting Up the Display Environment

pfPipe

pfChannel 0

pfChannel 1

w
Windo

pfPipe

pfScene

w
Windo

pfPipe

Scene graph

Display system

pfChannel 0

Figure 2-1

pfChannel 1

From Scene Graph to Visual Display

007-1680-080

Using Pipes

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:
•

“Using Pipes” on page 21

•

“Using Channels” on page 26

•

“Controlling the Video Output” on page 34

•

“Using Multiple Channels” on page 35

•

“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

007-1680-080

2: Setting Up the Display Environment

input to the draw process that performs the actual graphics library actions required to
draw the image.

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.

007-1680-080

Using Pipes

Application

Pipeline 0

Figure 2-3

Scene

Pipeline 1

Traversal/Cull

Draw

Traversal/Cull

Draw

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.

007-1680-080

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.

007-1680-080

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);

007-1680-080

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

007-1680-080

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.

007-1680-080

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.

007-1680-080

Using Channels

Horizontal FOV
x

Top

Far

Left
t

igh

Near

e
Lin

of s

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

007-1680-080

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

007-1680-080

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.

007-1680-080

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();

007-1680-080

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)
{

007-1680-080

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.

007-1680-080

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 14,
“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

007-1680-080

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.

007-1680-080

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

007-1680-080

Single-Channel and Multiple-Channel Display

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.

007-1680-080

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).
*/

007-1680-080

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.

007-1680-080

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.

007-1680-080

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];
...

007-1680-080

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

007-1680-080

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 14, “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 22, “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

007-1680-080

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);

007-1680-080

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);

007-1680-080

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

007-1680-080

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.

007-1680-080

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.

007-1680-080

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.

007-1680-080

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

007-1680-080

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:

New

Create and return a handle to a new node.

Get

Get node attributes.

Set

Set node attributes.

007-1680-080

Nodes

Find

Find a node based on its name.

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

007-1680-080

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

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.

007-1680-080

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.

007-1680-080

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.

007-1680-080

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

Figure 3-2

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

007-1680-080

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.

Root

D1
P2

Dynamic
coordinate
system

B
C

Figure 3-3

D2
A

Leaf

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.

007-1680-080

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);

007-1680-080

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);

007-1680-080

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

007-1680-080

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.

007-1680-080

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 16, “Dynamic Data.”
Table 3-4

007-1680-080

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.

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.

007-1680-080

Node Types

Eye

Car

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.

007-1680-080

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);
}

007-1680-080

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

007-1680-080

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.

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. */

007-1680-080

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);
...

007-1680-080

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 17, “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.

007-1680-080

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

007-1680-080

3: Nodes and Node Types

state and geometry and are described in the section, “Geometry Sets” 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,
...

muffler);
frame);
seats);
tires);

007-1680-080

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);

007-1680-080

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
#include
#include
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 */

007-1680-080

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);

007-1680-080

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)

007-1680-080

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);
}

007-1680-080

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

007-1680-080

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

pfSegSet

Global
Activation

pfFrame()
pfSync()

pfFrame()

pfFrame()
pfNodeIsectSegs(), pfChanNodeIsectSegs()

pfAppFrame()
Global
Callbacks

pfChanTravFunc()

pfIsectFunc()

Activation
within
Callback

pfApp()

pfCull()

pfDraw()

pfFrame()
pfNodeIsectSegs(), pfChanNodeIsectSegs()

Path
Activation

N/A

pfCullPath()

N/A

Modes

pfChanTravMode()

pfSegSet (also
discriminator
callback)

Node
Callbacks

pfNodeTravFuncs()

Traverser
Masks

pfChanTravMask()

pfSegSet mask

Traversee
Masks

pfNodeTravMask()

pfNodeTravMask()
pfGSetIsectMask()

007-1680-080

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

007-1680-080

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()).

007-1680-080

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;

007-1680-080

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.

007-1680-080

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.

007-1680-080

4: Database Traversal

PFIS_FALSE

PFIS_ALL_IN

PFIS_TRUE

Figure 4-1

Culling to the Frustum

007-1680-080

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.

007-1680-080

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.

007-1680-080

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.

007-1680-080

4: Database Traversal

Board

Pyramids

Cones

Spheres

Cubes

Board

Tile 1

Figure 4-3

Tile 2

Tile 3

Tile 4

Tile 5

Tile 6

Tile 7

Tile 8

Tile 9

How to Partition a Database for Maximum Efficiency

007-1680-080

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 21). 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

007-1680-080

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 9,
“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.

007-1680-080

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:

007-1680-080

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 115 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 9), 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

007-1680-080

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).

007-1680-080

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

007-1680-080

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
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.
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.

007-1680-080

103

4: Database Traversal

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.
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().

104

007-1680-080

Controlling and Customizing Traversals

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 108. 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:
•

Test instructions

•

Assign instructions

•

Add-bin instructions

•

Jump instructions

•

Return instruction

Test Instructions
Table 4-2 describes the test instructions.
Table 4-2

007-1680-080

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.

105

4: Database Traversal

Table 4-2

Test Instructions (continued)

Instruction

Description

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

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.

PFCULLPG_ASSIGN_BIN_FALSE b

Assigns bin b to the pfGeoSet if the result of the last
polytope 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.

106

007-1680-080

Controlling and Customizing Traversals

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

007-1680-080

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.

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.

107

4: Database Traversal

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.
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.

108

007-1680-080

Controlling and Customizing Traversals

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

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.

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;
}

007-1680-080

109

4: Database Traversal

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.
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.

110

007-1680-080

Controlling and Customizing Traversals

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)

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.

007-1680-080

111

4: Database Traversal

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.

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

112

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.

007-1680-080

Controlling and Customizing Traversals

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.
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().

007-1680-080

113

4: Database Traversal

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 9). 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;
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;

114

007-1680-080

Process Callbacks

}
/* 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;
}

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.

007-1680-080

115

4: Database Traversal

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)
{
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.
*/

116

007-1680-080

Process Callbacks

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);
}

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

007-1680-080

117

4: Database Traversal

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.
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);

118

007-1680-080

Process Callbacks

/* 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)
{
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

007-1680-080

119

4: Database Traversal

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.

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

120

007-1680-080

Intersection Traversal

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

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

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

007-1680-080

121

4: Database Traversal

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)

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

122

007-1680-080

Intersection Traversal

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
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.

007-1680-080

123

4: Database Traversal

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.
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

124

007-1680-080

Intersection Traversal

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.

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 125.
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 21, “Performance Tuning and Debugging”).

007-1680-080

125

4: Database Traversal

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().
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.

126

007-1680-080

Intersection Traversal

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.

Figure 4-4

007-1680-080

Intersection Methods

127

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.

007-1680-080

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 21, “Performance
Tuning and Debugging.”

130

007-1680-080

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.

007-1680-080

131

5: Frame and Load Control

Refresh count
modulo three

Overrun

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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. */

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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 7-3. 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

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.

007-1680-080

145

5: Frame and Load Control

This 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

007-1680-080

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 17, “Active Surface Definition.”

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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);

007-1680-080

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

007-1680-080

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

ter
Fil D)
s
s
O
e
Str Set L
(
sal
ver D)
a
r
O
T L
e
(Us
g

rin

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

007-1680-080

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

007-1680-080

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:

007-1680-080

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_CULL

156

007-1680-080

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).

007-1680-080

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 16, “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

007-1680-080

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

007-1680-080

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

007-1680-080

Successful Multiprocessing with OpenGL Performer

Boxes represent the functional stages and are labeled as follows:

007-1680-080

Application process for the nth frame

Cull process for the nth frame

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

PFMPAPPCULLDRAW

PFMPAPPCULL_DRAW
P1

PFMP_APP_CULLDRAW

PFMP_APP_CULL_DRAW P1

Start

PFMP_APP_CULL0DRAW P1

C1
D0

Frame 0

C2
D1

Frame 1

C3
D2

Frame 2

Frame 3

Frame 4

Time

Figure 5-6

162

Multiprocessing Models

007-1680-080

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

007-1680-080

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

007-1680-080

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.

007-1680-080

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 15). Configuration routines do not take effect until
pfConfig() is called. These are the configuration routines:
–

pfMultipipe()

–

pfMultiprocess()

–

pfMultithread()

–

pfHyperpipe()

007-1680-080

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 9.) 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:

•

007-1680-080

–

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.

007-1680-080

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

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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.

007-1680-080

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.

007-1680-080

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;

007-1680-080

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

007-1680-080

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).

007-1680-080

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.

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

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().

007-1680-080

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.

007-1680-080

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.

007-1680-080

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.

007-1680-080

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

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

007-1680-080

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

Use 2D texture

PFVFOG_FLAG_FORCE_2D_TEXTURE

Force patchy fog passes

PFVFOG_FLAG_FORCE_PATCHY_PASS

Self-shadowing

PFVFOG_FLAG_SELF_SHADOWING

Darken objects

PFVFOG_FLAG_DARKEN_OBJECTS

Filter color

PFVFOG_FLAG_FOG_FILTER

Faster patchy fog

PFVFOG_FLAG_FASTER_PATCHY_FOG

No object in fog

PFVFOG_FLAG_NO_OBJECT_IN_FOG

1D texture on surface

PFVFOG_FLAG_PATCHY_FOG_1DTEXTURE

Separate node bins

PFVFOG_FLAG_SEPARATE_NODE_BINS

Screen-bounding rectangle

PFVFOG_FLAG_SCREEN_BOUNDING_RECT

Draw nodes separately

PFVFOG_FLAG_DRAW_NODES_SEPARATELY

User-defined texture

PFVFOG_FLAG_USER_PATCHY_FOG_TEXTURE

Use cull programs

PFVFOG_FLAG_USE_CULL_PROGRAM

PFVFOG_FLAG_LAYERED_PATCHY_FOG

PFVFOG_FLAG_LIGHT_SHAFT

General

Layered fog

Patchy fog

Layered patchy fog
Use layered patchy fog
Light shafts
Light shaft

190

007-1680-080

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

007-1680-080

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

007-1680-080

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:
•

007-1680-080

“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

007-1680-080

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

007-1680-080

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.

007-1680-080

Patchy Fog and Layered Fog

007-1680-080

•

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”

007-1680-080

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);

007-1680-080

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.

007-1680-080

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

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.

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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 ]

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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.

216

007-1680-080

Image-Based Rendering

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
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.

007-1680-080

217

6: Creating Visual Effects

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,
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.

218

007-1680-080

Image-Based Rendering

•

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.

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.

007-1680-080

219

6: Creating Visual Effects

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.

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.

007-1680-080

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

•

Preexisting file-loading utilities

•

Several utility functions in the OpenGL Performer database utility library that can
make the process of database conversion easier for you

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

007-1680-080

Database-Importer Source Directories

Directory Name

Directory Contents

libpfdu

General database processing tools and utilities.

libpfdb

Load, convert, and store specific database formats..

libpfutil

Additional utility functions.

221

7: Importing Databases

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

007-1680-080

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

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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.

007-1680-080

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 9, 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

007-1680-080

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:

007-1680-080

•

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

007-1680-080

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)

007-1680-080

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;inumVerts;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

007-1680-080

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)
{

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

007-1680-080

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

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

Simple OpenGL Performer data format

irtp

AAI/Graphicon Interactive Real-Time PHIGS

SGI Open Inventor format (VRML 1.0 superset)

lsa

Lightscape Technologies ASCII radiosity data

lsb

Lightscape Technologies binary radiosity data

007-1680-080

Supported Database Formats

007-1680-080

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

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

007-1680-080

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];

007-1680-080

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

007-1680-080

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 <

007-1680-080

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

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

007-1680-080

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?

007-1680-080

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

007-1680-080

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:

007-1680-080

■

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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:

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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
...

007-1680-080

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

007-1680-080

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:

007-1680-080

•

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

007-1680-080

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

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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:

007-1680-080

•

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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

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

007-1680-080

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.

007-1680-080

277

7: Importing Databases

The layout of an SGO file is the following:

...

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

Independent quadrilaterals

OBJ_TRILIST

Independent triangles

OBJ_TRIMESH

Triangle mesh

OBJ_END

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

007-1680-080

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

Begin a triangle strip.

OP_SWAPTMESH

Exchange old vertices.

OP_ENDBGNTMESH

End then begin a strip.

OP_ENDTMESH

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.

007-1680-080

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

007-1680-080

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.

007-1680-080

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

007-1680-080

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.

007-1680-080

283

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 [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

Description of Supported Formats

–
•

Sn Tn are the nth texture coordinates

3D Triangle mesh with vertex normals and optional texture coordinates
tmeshn [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

007-1680-080

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

007-1680-080

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,

007-1680-080

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
operate on. Loaders can generate their scene graphics procedurally based on simple
parameters specified in the command string.

288

007-1680-080

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 app 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.

Pseudo loaders should define pfdLoadNeededDSOs_EXT() for the following:

007-1680-080

•

Preinitializing DSOs

•

Loading other special files

•

Performing additional initialization, such as class initialization, that should happen
before pfConfig()

289

Chapter 8

8. Geometry

All libpr geometry is defined by modular units that employ a flexible specification
method. These basic groups of geometric primitives are termed pfGeoSets.

Geometry Sets
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 299 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.
Note: libpf applications can include pfGeoSets in the scene graph with the pfGeode
(Geometry Node).

007-1680-080

291

8: Geometry

Table 8-1 lists a subset of the routines that manipulate pfGeoSets.
Table 8-1

292

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.

pfQueryGSet()

Determine the number of triangles or vertices.

pfPrint()

Print the pfGeoSet contents.

007-1680-080

Geometry Sets

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

The number of primitive items in the pfGeoSet, as set by
pfGSetNumPrims().

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

007-1680-080

293

8: Geometry

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
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:

294

•

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.

007-1680-080

Geometry Sets

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
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.

007-1680-080

295

8: Geometry

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.
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,

296

007-1680-080

Geometry Sets

and D, and the geometry resulting from different primitive types. This example uses
index lists that index the coordinate array.

007-1680-080

297

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

Independent Quadrilaterals
triangles

Triangle strips

Polygons

Geometry

Index list

Figure 8-1

298

...

Primitives and Connectivity

007-1680-080

Geometry Sets

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 9, “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

007-1680-080

299

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 16, “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

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

ST
2: S
T

1: S
T
2: S
T

0: S
T
1: S
T
2: S
T

dxc

oor

300

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

a
rm
Nodex
in

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

007-1680-080

Geometry Sets

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

PFGS_PER_PRIM

Yes

PFGS_PER_VERTEX

Yes

PFGS_OFF

Yes

Attribute lists, index lists, and binding types are all set by pfGSetAttr().

007-1680-080

301

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.

302

007-1680-080

Geometry Sets

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.

007-1680-080

303

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.

304

007-1680-080

3D Text

Table 8-1 on page 292 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 19 for more information on that function.

3D Text
In addition to the pfGeoSet, 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.
Table 8-5 lists some routines that are used with a pfFont.
Table 8-5

007-1680-080

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.

305

8: Geometry

pfFont Routines (continued)

Table 8-5
Routine

Description

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;isetVirtualLODOffset(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

007-1680-080

431

12: 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().

432

007-1680-080

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.

007-1680-080

433

12: 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.

434

007-1680-080

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

007-1680-080

435

12: 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 read 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 customer 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

436

007-1680-080

Using Cliptextures

For hints on when and how to use custom read functions, see the customizing read
functions in “Custom Read Functions” on page 448.

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 values.
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.

007-1680-080

437

12: ClipTextures

Insets

Figure 12-12

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.

438

007-1680-080

Using Cliptextures

DTR does this examination starting with the coarsest clipped level. If the current level is
complete, it marks that level as visible and repeats the process for the next finest level,
until it gets to the top level or finds an incomplete level.
Building Insets

To create an inset, assume you have a cliptexture that is complete at a coarse level. You
choose an area of the clipmap that you would like to have visible at some finer level. In
order to make that area available, you have to provide texel data in that area for each
level from the coarse complete level to the finer level you want to show.
When the clip region is completely enclosed by the finer level data, DTR checks all the
levels from the pyramid level on up, and allows the finer level to be shown in that area.
(The pyramid levels are always complete; see Figure 12-1.)
Because DTR works from the bottom up (coarser to finer levels), an inset area must have
texel data available from the finest level all the way down to the pyramid level. If a level’s
clip region is missing or incomplete, DTR does not allow the image to sharpen up to that
level; the inset gets blurry.
Inset Boundaries

When the clipcenter is set such that the clip region is completely enclosed by the insetted
area, a properly constructed inset is sharp, using the finer resolution texel data. But what
happens when the clip region only partially covers an inset? In that case, DTR does not
sharpen up beyond the finest complete level, and the clip region gets blurry, including
the part of the clip region covered by the inset. Remember, the clip region only sharpens
to the finest level that is complete within the clip region.
This bluriness may not be a problem. If you know that the application moves far enough
away from the terrain before the clip region crosses an inset border, MIPmapping uses
the coarser texture levels before DTR forces the texture to use them. Sometimes, the
application would like a static boundary between the inset and the surrounding coarser
data, even when crossing an inset boundary close to the textured geometry.
Supersampled Data

Getting a static inset border requires creating a boundary of supersampled data around
the inset. This means creating texel data for the insetted levels that is deliberately made

007-1680-080

439

12: ClipTextures

as blurry as the surrounding base level. This border of blurry data must be at least as
wide as the clip region to ensure a smooth transition.

Inset
Supersampled boundary
Coarse level

Figure 12-13

Supersampled Inset Boundary

How does this work? When the clip region crosses an inset border, it starts to cover the
boundary region. Since the boundary region has the same levels filled in as the inset
region, DTR still sees complete data up to the same finer level. The texel data of the
border has been blurred, however, so it looks like the coarser base level. This allows a
hard transition between the finer inset data and the surrounding coarse data. Since DTR
still sees complete levels, as you move away from the inset, it does not suddenly blur.
Since the boundary data is at least as wide as the clip region, the inset boundary has
moved out of the clip region before the clip region hits the far edge of the boundary
region. At this point, DTR blurs down to the coarser base level, since the finer data is no
longer incomplete, but there is no visual change, since the boundary texels were blurry
already.
Note: Supersampled borders do not guarantee a seamless transition between insets and
their surroundings, only that the inset region does not suddenly blur or sharpen as the
clip region crosses the inset border. Seamless transitions only happen if the application is
careful to get far enough from the clipmapped terrain to already be using the coarse
levels before crossing the inset border.

440

007-1680-080

Using Cliptextures

Multiple Insets

Insets do not have to be any particular size or shape, although they are usually multiples
of the tile size, and at least as big as the clip region so they can be viewed without a seam
of bluriness. Typically, insets are designed so when the application is close enough to
view the finer inset levels, the clip region is already completely enclosed by the inset
region.
Keep in mind that insets are not the same size on each level, since texels from coarser
levels cover more geometry. If an inset is not a multiple of the tile size at a given level,
the tile has to be partially inset data, and partially supersampled data, or all fine data,
since DTR does not work with partial tiles.

Estimating Cliptexture Memory Usage
Because cliptextures are a voracious consumer of system and texture memory, it is
important to accurately predict the system resources required to run a cliptexture
application. It is better to customize your application than to rely on the cliptexture auto
resizing feature, since auto-resizing does not take into account multiple cliptextures or
pfTextures in your application.
Cliptextures use both system memory (texel caching, read queue elements as well object
overhead) and texture memory. The following estimation ignores the smaller
contributors to system memory overhead and concentrates on image cache consumption
of system memory for mem regions and tex regions in texture memory.
System Memory Estimation

The following values are required to estimate system memory requirements:
•

Size of clipmap level 0

•

Clip size (in texels)

•

Tile size (in texels; assuming tile size is the same for all levels)

•

Texel size in bytes

•

Whether the tiles have high disk latency

Given these values, compute the estimate for system memory requirements using the
following procedure:

007-1680-080

441

12: ClipTextures

Round up clip size to even multiple of tile size in each dimension.

2. Divide each dimension by tile size in that dimension.
3. Add 2 tiles to each dimension for a tile boundary of 1.
If there is high latency downloading, such as reading tiles over the network or
decompressing tiles, add 4 tiles per dimension, giving a tile boundary of 2.
You now have the number of tiles in each dimension per clipped level.
4. Multiply each tile number in each dimension by the corresponding tile size in that
dimension.
You now have the number of texels in each dimension.
5. Multiply the texel dimensions together.
6. Scale by the size of each texel.
7. Add in the fixed cost of image cache structs.
You now have the system memory cost in bytes for each clipped level.
8. Treat each level bigger than clip size as clipped. Add 4/3 of the clip size scaled by
the texel size for the pyramid levels.
This estimate is a bit too conservative, since the lowest clipped levels may exceed
the entire level size with a border of two tiles. It is a function of tile size and clip
size.
9. Scale the clipped level size by the number of clipped levels.
Example Estimating System Memory Requirements

The example uses the following values in its estimation:
•

2M top level

•

1K clip size

•

512 tile size (everything square)

•

1 byte texel size (LMV example)

Example 12-1 estimates system memory requirements using the preceding procedure.

442

007-1680-080

Using Cliptextures

Example 12-1

Estimating System Memory Requirements

No-op: 1K, 1K (for both s and t).

2. 1K/512 = 2, 2 (for both s and t).
3. 2+4 = 6, 6 (for both s and t; high latency on tile download)
4. 6 * 512 = 3K, 3K (for both s and t)
5. 3K * 3K = 9M
6. 9M * 1 = 9M
7. 9M * 10 = 90M (for 2M -> 4K levels) + 2M (for 2K level) = 92M
8. 4/3 * 1K * 1K= 1.3M
9. Total Size 92M + 1.3M = 93.3M
Texture Memory Estimation
The following values are required to estimate texture memory requirements:
•

clip size (in texels)

•

whether the clipmap is virtual or non-virtual

•

number of levels in use (if less than 16)

Given these values, you can compute the estimate for texture memory requirements
using one of the following guidelines:
•

If the clipmap is virtual, multiply the number of levels by the square of the clip size.

•

If the clipmap is non-virtual, do the following:
1.

Multiply the number of levels bigger than the clip size by the square of the clip
size.

Add 4/3 times the clip size squared.

There is, however, a further complication involved in accurately estimating texture
memory requirements. The following subsection describes it.
Texture Memory Usage: A Further Complication

InfiniteReality rendering boards come with either 16 or 64 megabytes of texture memory.
Unfortunately, you cannot just use the texture memory any way you want. The texture

007-1680-080

443

12: ClipTextures

memory is divided into two equal banks. Each adjacent Mipmap level, clipmapped or
not, must be placed in opposite banks. This ends up restricting the amount of texture
memory available for clipmapping.
A further restriction is that texture formats can take up 16 bits or 32 bits of data per texel,
but nothing in between. This means an 888 RGB format takes up 32 bits per texel, just like
an 8888 RGBA format.
To give you an example of this restriction, consider an example using RGB texel data, 8
bits per component, and a clip size of 2048 by 2048. The largest level is 8K by 8K. The
system has an RM board with 64M of texture memory; so, it would seem that there is
plenty room, but the following calculation shows otherwise:
1.

The cliptexture is non-virtual; so, the total texture memory requirement is the clip
size times the number of clipped levels plus 4/3 of the pyramid.

2. 4K, 4K levels are clipped to 2K X 2K: RGB, 8 bits per channel 4 bytes (not 3) per texel
times 2K X 2K = 4M of texels per level.
3. There is 6M of texture memory per clipped level.
4. So, that is 32M of texture memory.
5. 2K and below is the pyramid; so, 4/3 of 16M = 21-1/3M.
6. The total is 53-1/3M of texture memory.
Unfortnately, the texture does not fit into texture memory because of the following:
1.

64M of texture memory means two 32M banks.

2. Each level must be in the opposite 32M bank.
3. Consequently, 8K level becomes 16M in bank 0 (16M left).
4. At the 4K level, 16M goes into bank 1 (16M left).
5. At the 2K level, 16M goes into bank 0 (0M left).
6. At the 1K level, 8M goes into bank 1 (8M left).
7. At the 512 level, there is no room in bank 0
The best you could do is to have only one clipped level, as follows:
1.

At the 4K level,16M goes into bank 0 (16M left).

2. At the 2K level, 16M goes into bank 1 (16M left).

444

007-1680-080

Using Cliptextures

3. At the 1K level, 8M goes into bank 0 (8M left).
4. At the 512 level, 4M goes into bank 1 (12M left).
5. At the 256 level, 2M goes into bank 0 (6M left) and so on.
Probably a better solution would be to use the 5551 format RGBA texels, which only use
16 bits per texel, allowing more levels, as follows:
1.

At the 32K level, 8M goes into bank 0 (24M left).

2. At the 16K level, 8M goes into bank 1 (24M left).
3. At the 8K level, 8M goes into bank 0 (16M left).
4. At the 4K level, 8M goes into bank 1 (16M left).
5. At the 2K level, 8M goes into bank 0 (8M left).
6. At the 1K level, 4M goes into bank 1 (12M left).
7. At the 512 level, 2M goes into bank 0 (6M left).
8. At the 256 level, 1M goes into bank 1 (11M left) and so on.
You can get a lot more mileage out of smaller texel formats than fewer levels. This
becomes even more true for RMs with only 16M of texture memory.

Using Cliptextures in Multipipe Applications
OpenGL Performer provides good support for multipipe cliptextures, allowing
applications to ignore many of the differences between single pipe and multipipe
operations. The primary issue in multipipe applications is knowing when to make
master/slave cliptextures, what parameters should be shared, and when and how to
create separate centers in different master and slave cliptextures.
When to Make Master/Slave Cliptexture Groups

When it is possible, it is desirable to make master/slave cliptexture groups in multipipe
applications. Master/slave groups share the same mem regions and disk access
bandwidth, reducing the load on the system. Masters and slave can also take advantage
of sharegroups, automating some of the work of synchronizing slave cliptextures with
their masters.

007-1680-080

445

12: ClipTextures

Master and slave cliptextures do not work when the different cliptextures represent
different textures. Separate cliptextures must be created for each texture that has to be
displayed. Another condition that prevents using master/slave cliptexture groups is
when the center for the cliptexture in each pipe is completely independent. Master and
slave cliptextures assume that the tex region for each cliptexture is always completely
enclosed by the shared mem region. If that assumption is violated, the cliptexture data
will be invalid for the parts of the tex region outside of the shared mem region, and the
cilptextures will print error messages.
Slave Cliptextures with Offset Centers

There is no reason for the slave cliptextures to have the same centers as the master, as
long as the tex regions always stay within the mem region. Sometimes it is desirable for
an application to have views with slightly different viewpoints for each pipe. This can be
done by turning of clipcenter sharing and having the application set a slave’s center
directly. The application is responsible for keeping the tex region inside the master’s
mem region at all times. The position of the master’s center determines the groups mem
region, so the master’s center can not be offset from the center of the mem region.
Tex region
Mem region

Offset tex region
Offset tex region

ste

r
Sla

ve
Sla

Figure 12-14

Offset Slave Tex Regions

Virtualizing Cliptextures
Virtual cliptextures are one of the most challenging features to use in an OpenGL
Performer application. Cliptextures themselves are challenging enough, since they tie
together functionality in the scene graph, pfPipes and pfChannels. Virtual cliptextures

446

007-1680-080

Using Cliptextures

add an additional level of complication, since they require that the application segment
the terrain, estimate the texture LOD levels needed for each terrain segment based on the
current eyepoint, and update the virtual cliptexture parameters appropriately.
Some tips to consider when working with virtual cliptextures:
•

Get your application working with a non-virtual cliptexture first. It is hard to
separate virtual cliptexture problems form basic cliptexture configuration problems
(which are usually easier to fix).

•

Start off with a single segment and as little complexity in the application as possible;
then get that working. It makes debugging much simpler.

•

When in doubt, print it out. For debugging purposes, a well placed set of pfNotify()
statements can be really helpful. It is also useful to set up a canonical scene, where
you know what parameters should be generated, then compare them against what
the program does.

•

Take maximum advantage of sample code and utilities. Try to re-use some of the
example code provided by OpenGL Performer. Using (or just reading) through the
loaders, example programs, and utilities listed in this section “Cliptexture Sample
Code” on page 448 can save you hours of work.

Customizing Load Control
Setting texload time, multiple cliptexture load control, normally, bandwidth from disk to
system memory is optimized by image cache configuration. You can use the streams
feature in image cache configuration files to maximize bandwidth by configuring the
system with as more disks and disk controllers, and copying the texture data files over
multiple disks. By using the streams feature in image cache configuration files, you can
then parallelize the disk downloads over separate disks and disk controllers. You can
also stripe disks to increase disk download bandwidth.
Texture memory bandwidth is more a matter of careful rationing of DRAW process time
each frame. You will have to ration between sending geometry and texture data to the
graphics pipeline. You can adjust the cliptexture’s texload time to minimize idle time in
the draw process. DTR computes texture download time using a cost table to estimate
what the download updates will cost each frame. It is a good estimate but not a perfect
one. You may have to build in some time margin in the DRAW process to avoid dropping
frames.

007-1680-080

447

12: ClipTextures

Download time setting can be as simple as finding the minimum time that still allows the
cliptexture to completely sharpened when the clipcenter is stationary. If the texload time
is too small, a cliptexture can get “stuck” and never sharpen beyond a given level.
If you are a sophisticated user, you might consider adjusting the invalid border as well,
which will change the effective clip size of the cliptexture. This could be used to allow a
new level to be gradually loaded over a number of frames, trading off a finer visible level
against a smaller effective clip region.

Custom Read Functions
Custom read functions allow the application to control what happens when texture data
is downloaded from disk. Operations include: texture data decompression, texture data
image processing, signaling an update of a disk cache from tape, etc. As described in the
API section, replacing the read function is fairly easy to do in OpenGL Performer.
There is one caveat: increasing the overhead of read functions can have undesirable
consequences. The latency of read operations determines the minimum size of the image
cache mem regions, since they need to lookahead to compensate for high latency reads.
Low bandwidth read operations effect how fast the center can move before DTR must
blur down to coarser levels.
One way around these problems is to implement a lookahead disk cache. Rather than
have the read function decompress files, have it read files from a cache of files that have
already been decompressed by an asynchronous process. The read function can signal
the other process to decompress more files as it gets close to the edge of the cache.
Texture data image processing, since it is relatively fast compared to disk reads, usually
does not require such elaborate measures. Changing the read function can be done in
conjunction with modifying the image cache mem region data to ensure that all data read
from disk is processed. A good example of this technique is the gridify feature, described
in “Invalidating Cliptextures” on page 428.

Cliptexture Sample Code
The best way to learn to use cliptextures is to work from existing code. OpenGL
Performer has a number of demo programs, test programs, loader code, and utilities,
with different levels of sophistication:

448

007-1680-080

Using Cliptextures

Test and Demo Programs

•

/usr/share/Performer/src/pguide/libpr/C/icache.c for IRIX and Linux
and %PFROOT%/Src/pguide/libpr/C/icache.c for Microsoft Windows
This is a simple libpr-style program that uses an image cache and the texture
transform matrix to scroll texture across a polygon.

•

/usr/share/Performer/src/pguide/libpr/C/icache_mwin.c for IRIX
and Linux and %PFROOT%/Src/pguide/libpr/C/icache_mwin.c for
Microsoft Windows
This program is a multiwindow version of icache; it uses master and slave image
caches. Note that the tex regions of the slaves image caches are offset from the
master.

•

/usr/share/Performer/src/pguide/libpr/C/cliptex.c for IRIX and
Linux and %PFROOT%/Src/pguide/libpr/C/cliptex.c for Microsoft
Windows
This is a simple libpr-style program that uses a cliptexture to display a bird’s eye
view of data. On a system that supports cliptexturing, it moves the clipcenter,
allowing the user to see the rectangle of texture resolution move as the center
translates between opposite diagonals.

•

/usr/share/Performer/src/pguide/libpr/C/cliptex_mwin.c for IRIX
and Linux and %PFROOT%/Src/pguide/libpr/C/cliptex_mwin.c for
Microsoft Windows
This is a multiwindow version of cliptex. It uses master and slave cliptextures.

•

/usr/share/Performer/src/pguide/libpf/C/cliptex.c for IRIX and
Linux and %PFROOT%/Src/pguide/libpf/C/cliptex.c for Microsoft
Windows
This is a libpf implementation using cliptextures. Rather than use clipcenter
nodes, it uses a simple centering mechanism based on the x and y coordinates of the
channel’s viewpoint.

•

/usr/share/Performer/src/pguide/libpf/C/virtcliptex.c for IRIX
and Linux and %PFROOT%/Src/pguide/libpr/C/virtcliptex.c for
Microsoft Windows
This is the virtual cliptexture version of the cliptex program. It will take any size
cliptexture and virtualize it. It divides a flat terrain in the x,y plane into a
rectangular grid, attaching geodes with callbacks to each grid square. The callback
calculates the virtual cliptexture parameters and applies them.

007-1680-080

449

OpenGL Performer Cliptexture Applications

•

/usr/share/Performer/src/app/sample/C/perfly.c for IRIX and Linux
and %PFROOT%/Src/app/sample/C/perfly.c for Microsoft Windows
This sample application supports cliptextured scene graphs. It assumes that the
loaders will do the basic configuration, but it does the post loading cliptexture
configuration, creating pfMPClipTextures and attaching them to pipes. It assumes
that clipcenter nodes are in the scene graph and that they will do the centering
work. Pressing g will toggle the gridify feature.

•

/usr/share/Performer/src/app/sample/C/clipfly.c for IRIX and Linux
and %PFROOT%/Src/app/sample/C/clipfly.c for Microsoft Windows
This is the cliptexture version of the perfly sample application. It contains extra
interface sliders and buttons that allow you to control many more cliptexture
parameters.

Loaders that Support Cliptextures

•

/usr/share/Performer/src/lib/libpfdb/libpfim/pfim.c for IRIX and
Linux and %PFROOT%/Src/lib/libpfdb/libpfim/pfim.c for Microsoft
Windows
This is the OpenGL Performer example loader. It contains simple tokens to create
clipcenter nodes and cliptextures. It is a good starting point for cliptexture
configuration in the loader.

•

/usr/share/Performer/src/lib/libpfdb/libpfct/pfct.C for IRIX and
Linux and %PFROOT%/Src/lib/libpfdb/libpfct/pfct.C for Microsoft
Windows
This is a more sophisticated cliptexture sample loader. It automatically creates some
simple cliptextured geometry, and supports virtual cliptextures, providing the
geometry segmentation and scene graph callbacks.

•

/usr/share/Performer/src/lib/libpfdb/libpfvct/pfvct.C for IRIX
and Linux and %PFROOT%/Src/lib/libpfdb/libpfvct/pfvct.C for
Microsoft Windows
This is a very simple pseudo-loader that will adjust a cliptextures parameters so it
can be used as a virtual cliptexture, even if its dimensions are smaller than 32K by
32K.

•

/usr/share/Performer/src/lib/libpfspherepatch/pfspherepatch.C
for IRIX and Linux and

Using Cliptextures

%PFROOT%/Src/lib/libpfspherepatch/pfspherepatch.C for Microsoft
Windows
This is a cliptexture loader designed to apply a cliptexture to a sphere. It is a good
example of a specific cliptexture application.
Cliptexture Utility Code

•

/usr/share/Performer/src/lib/libpfutil/gridify.C for IRIX and
Linux and %PFROOT%/Src/lib/libpfutil/gridify.C for Microsoft Windows
The gridify functionality illustrates techniques for dynamically modifying
cliptexture data and illustrates how to replace the cliptexture’s read function.

•

/usr/share/Performer/src/lib/libpfutil/trav.c for IRIX and Linux
and %PFROOT%/Src/lib/libpfutil/trav.c for Microsoft Windows
The function pfuFindClipTextures() illustrates how to traverse the scene graph to
find cliptextures.

•

/usr/share/Performer/src/lib/libpfutil/pfuClipCenterNode.C for
IRIX and Linux and %PFROOT%/Src/lib/libpfutil/pfuClipCenterNode.C
for Microsoft Windows
This code defines the clipcenter node and contains example code for creating
customized clipcenters by subclassing.

•

/usr/share/Performer/src/lib/libpfutil/clipcenter.c for IRIX and
Linux and %PFROOT%/Src/lib/libpfutil/clipcenter.c for Microsoft
Windows
The pfuProcessClipCenters() and pfuProcessClipCentersWithChannel() functions
illustrate how to use clipcenters in an application.

•

/usr/share/Performer/src/lib/libpfutil/cliptexture.c for IRIX
and Linux and %PFROOT%/Src/lib/libpfutil/cliptexture.c for Microsoft
Windows
The pfuAddMPClipTextureToPipes() and pfuAddMPClipTexturesToPipes()
functions illustrate how to work with cliptextures and pipes.

007-1680-080

451

Chapter 13

13. Windows

Rendering to the graphics hardware requires a window. A window is an allocated area
of the screen with associated framebuffer resources.
For IRIX and Linux, the X window system manages the use of shared resources among
the different windows. Windows can be requested directly from an X window server. By
use of the GLX extension, OpenGL Performer-based OpenGL graphics contexts can
render into X windows.
For Microsoft Windows, HWNDs and OpenGL/WGL contexts are used.
This chapter describes how to create, configure, manipulate, and communicate with a
window using pfWindow in OpenGL Performer. The extended libpf object
pfPipeWindow, based on pfWindow, is also mentioned in this chapter as the two objects
share much functionality. Likewise, for good understanding of windowing issues, read
the next chapter, Chapter 14, “pfPipeWindows and pfPipeVideoChannels.”

pfWindows
The pfWindow object provides an efficient windowing interface between your
application and the X Window System or the Microsoft Windows GUI. pfWindows
typically create an X window (an HWND on Microsoft Windows) or can also be
configured to embed your rendering area within a window created with Motif® or other
windowing toolkits. libpr provides utilities to shield you from the differences between
the different types of windows and guide you in your dealings with the window system.
pfWindows also keep track of your graphics state: they include a pfState which is
automatically initialized when you open a window and switched for you when you
change windows. Simple libpr windowing support centers around the pfWindow. The
libpf windowing support utilizes a pfWindow as part of a pfPipeWindow.
OpenGL Performer automatically configures and initializes your window so that it will
be ready to start rendering efficiently. In the simplest case, pfWindows make creating a

007-1680-080

453

13: Windows

graphics application that can run on any SGI machine with OpenGL a snap. pfWindows
do not limit your ability to configure any part or all of your windowing environment
yourself; you can use the libpr pfWindows to manage your GL windows even if you
create and configure the actual windows yourself.

Creating a pfWindow
A pfWindow structure is created with pfNewWin(). It can then be immediately opened
with pfOpenWin(). Example 13-1 shows the most basic pfWindow operations in libpr
program: to open and clear a pfWindow and swap front and back color buffers.
Example 13-1

Opening a pfWindow

int main (void)
{
pfWindow *win;
/* Initialize Performer */
pfInit();
pfInitState(NULL);
/* Create and open a Window */
win = pfNewWin(NULL);
pfWinName(win, “Hello from OpenGL Performer”);
pfOpenWin();
/* Rendering loop */
while (1)
{
/* Clear to black and max depth */
pfClear(PFCL_COLOR | PFCL_DEPTH, NULL);
...
pfSwapWinBuffers(win);
}
}

The pfWindow in Example 13-1 will have the following configuration:
Window system interface
An OpenGL window using the OpenGL/X GLX interface on IRIX and
Linux or an OpenGL HWND on Microsoft Windows.

454

007-1680-080

Creating a pfWindow

Screen

On IRIX and Linux, the pfWindow will open a window on the screen
specified by the DISPLAY environment variable or else on screen 0. On
Microsoft Windows, screen 0 will always be used.

Position and size
The position and size will be undefined and the window will come up
as a rubber-band for the user to place and stretch.
Framebuffer configuration
The window will be double-buffered RGBA with depth and stencil
buffers allocated. The size of these buffers will depend on the available
resources of the current graphics hardware platform. pfWindows will
also have multisample buffers allocated if they are available on current
hardware platform.
libpr state

A pfState will be created and initialized with all modes disabled and no
attributes set.

Graphics state The pfWindow will be in RGBA color mode with subpixel vertex
positioning, depth testing and viewport clipping enabled. The viewing
projection will be a two-dimensional one-to-one orthographic mapping
from eye coordinates to window coordinates with distances to near and
far clipping planes -1 and 1, respectively. The model matrix will be the
current matrix and will be initialized to the identity matrix.
Typically, pfWindows go through a bit more initialization than that of Example 13-1. The
pfWindow type, set with pfWinType(), is a bitmask that selects the window system
interface and the type of rendering window. Table 13-1 lists the possible selectors that can
be ORed together for specification of the window type.
Table 13-1
PFWIN_TYPE_
Bitmask Token

007-1680-080

pfWinType() Tokens
Description

Window will be an X window on IRIX and Linux but will be an HWND on
Microsoft Windows. This is the default.

STATS

Window will have framebuffer resources to accommodate hardware statistics
modes. This type cannot be combined with PFWIN_TYPE_OVERLAY or
PFWIN_TYPE_NOPORT.

OVERLAY

Window will have only overlay planes for rendering. This type cannot be
combined with PFWIN_TYPE_STATS or PFWIN_TYPE_NOPORT.

455

13: Windows

Table 13-1

pfWinType() Tokens (continued)

PFWIN_TYPE_
Bitmask Token

Description

NOPORT

Window will have a graphics context but no physical window or graphics or
framebuffer rendering resources and will not be placed on the screen. This
token can not be used in combination with any other type token.

PBUFFER

The pfWindow drawable will be created as a pbuffer and will not be visible.

UNMANAGED

Other than select upon open, no unrequested window management operations
are done automatically on the pfWindow.

The selection of screen can be done explicitly with pfWinScreen(), or implicitly by
opening a connection to the window system using pfOpenScreen() with the desired
screen as the default screen. A window system connection can communicate with any
screen on the system; the default screen only determines the screen for windows that do
not have a screen explicitly set for them. Only one window system connection should be
opened for a process. See “Communicating with the Window System” later in this
section for details on efficient interaction with the window system.
The position and/or size, is set with pfWinOriginSize(). If the x and y components of the
origin are (-1), the window will open with position undefined for the user to place. If the
x or y components of the size are (-1), the window will open with both position and size
undefined (the default) for the user to place and stretch. On IRIX and Linux, the X
window manager may override negative origins and place the window at (0,0). If the
window is already opened when pfWinOriginSize() is called, the window will be
reconfigured to the specified origin and size upon the next pfSelectWin(). Similarly,
pfWinFullScreen() causes a window to open as full screen or to become full screen upon
the next call to pfSelectWin(). A full screen window will have its border automatically
removed so that the drawing area truly gets the full rendering surface. The routines for
querying the position and size work a bit differently than the pattern established by the
rest of libpr get and set pairs of routines. This is because a user may change the origin
or size independently of the program and under certain conditions, querying the true
current X window size and origin can be expensive. pfGetWinOrigin() and
pfGetWinSize() will always be fast and returns the last explicitly set origin and size,
such as by pfOpenWin(), pfWinOriginSize(), or pfWinFullScreen(). If the window
origin or size has been changed, but not through a pfWindow routine, the values
returned by pfGetWinOrigin() and pfGetWinSize() may not be correct.
pfGetWinCurOriginSize() returns an accurate size and origin relative to the pfWindow
parent. For X windows, note that it requires an expensive query to the X server and
should not be done in real-time situations. pfGetWinCurScreenOriginSize() returns the

456

007-1680-080

Configuring the Framebuffer of a pfWindow

size and the screen-relative origin of the pfWindow. As with pfGetWinCurOriginSize(),
this command will be quite expensive and is not recommended accept for rare use or
initialization purposes.
pfPipeWindows, discussed in Chapter 14, “pfPipeWindows and pfPipeVideoChannels,”
take advantage of the multiprocessed libpf environment to always be able to return an
accurate window size and origin relative to the window parent. However, even for
pfPipeWindows, getting a screen-relative origin can be an expensive operation.
Hint: Write programs that are window-relative and do not depend on knowing the
current exact location of a window relative to its parent or screen.

Configuring the Framebuffer of a pfWindow
OpenGL Performer provides a default framebuffer configurations for the current
graphics hardware platform for the standard window types: normal rendering, statistics
(stats), and overlay. You may want to define your own framebuffer configuration, such
as single-buffered, stereo, etc. You can use utilities in libpr to help you with this task,
or create your own framebuffer configuration structure with X utilities, or even create the
window yourself and apply it to the pfWindow. pfOpenWin() respects any specified
framebuffer configuration. Additionally, pfOpenWin() uses any window or graphics
context that is assigned to it and only creates what is undefined.
pfWinFBConfigAttrs() can be used to specify an array of framebuffer attribute tokens
listed in Table 13-2. The tokens correspond to OpenGL/X tokens. Note that if an attribute
array is specified, the tokens modify configuration with no attributes set, not the default
OpenGL Performer framebuffer configuration.

007-1680-080

457

13: Windows

Table 13-2

pfWinFBConfigAttrs() Tokens

PFFB_ Token

Value

Description

BUFFER_SIZE

integer > 0

The size of the color index buffer.

LEVEL

integer > 0

The color plane level:
normal color planes have level = 0
overlay color planes have level > 0
underlay color planes have level < 0
There may be only one or no levels for overlay and
underlay color planes on some graphics hardware
configurations.

RGBA

Boolean: true Use RGBA color planes (instead of color index).
if present

DOUBLEBUFFER

Boolean: true Use double-buffered color buffers.
if present

STEREO

Boolean: true Allocate left and right stereo color buffers (allocates
if present
back left and back right if DOUBLEBUFFER is specified.

AUX_BUFFER

integer > 0

Number of additional color buffers to allocate.

RED_SIZE
GREEN_SIZE
BLUE_SIZE
ALPHA_SIZE

integer > 0

Minimum number of bits color for components R, G,
and B will all be the same and be the maximum
specified. Alpha may be different.

DEPTH_SIZE

integer > 0

Number of bits in the depth buffer.

STENCIL

integer > 0

Number of bits allocated for stencil. One is used by
pfDecal rendering and three or four are used by the
hardware fill statistics in pfStats.

integer > 0
ACCUM_RED_SIZE
ACCUM_GREEN_SIZE
ACCUM_BLUE_SIZE
ACCUM_ALPHA_SIZE
USE_GL

458

Number of bits per RGBA component for the
accumulation color buffer.

Boolean: true Exists for historical reasons. Has no effect.
if present

007-1680-080

pfWindows and GL Windows

If you desire more control over the exact framebuffer configuration of your pfWindow,
you have several options. For OpenGL/X windows you can provide the appropriate
framebuffer description for the current GL operation to the pfWindow using
pfWinFBConfig(). X uses visuals to describe available framebuffer configurations.
XVisualInfo pointer with XGetVisualInfo() returns a list of all visuals on the system and
you can search through them to find the appropriate configuration. On IRIX-based
systems, OpenGL/X also uses GLXFBConfigSGIX to describe framebuffer
configurations. You can select either the visual or the GLXFBConfigSGIX for your
window and set it on the pfWindow with pfWinFBConfig(). pfGetWinFBConfig()
always returns the corresponding X visual.
libpr also offers utilities for creating framebuffer configurations (pfFBConfig)
independently of a pfWindow. pfChooseFBConfig() takes an attribute array of tokens
from Table 13-2 and will return a pfFBConfig structure that can be used with your
pfWindows, or with X Windows created outside of libpr, such as with Motif. You may
get back a framebuffer configuration that is better than the one you requested. OpenGL
Performer will give you back the maximum framebuffer configuration that meets your
request that will not add any serious performance degradations. There are specific
machine-dependent instances where, for performance reasons, we do limit the
framebuffer configuration. See the man page for pfChooseWinFBConfig() for the
specific details. The libpfutil utility pfuChooseFBConfig() in
/usr/share/Performer/src/lib/libpfutil/xwin.c (IRIX and Linux) and
%PFROOT%/Src/lib/libpfutil/xwin.c (Microsoft Windows) provides a limiting
framebuffer configuration selector, complete with source code.
You can use pfQuerySys() to query particular framebuffer resources in the current
hardware configuration and then use pfQueryWin() to query your resulting framebuffer
configuration.

pfWindows and GL Windows
Note: This is an advanced topic.
For IRIX and Linux, libpr allows you to use X window handles and OpenGL/X
graphics contexts to create your own windows and to set them on the pfWindow. These
handles can be assigned to the pfWindow with pfWinWSDrawable() or pfWinGLCxt().

007-1680-080

459

13: Windows

For Microsoft Windows, you can do the same, but you must use HWNDs (instead of X
window handles) and OpenGL/WGL graphics contexts.
pfOpenWin() will automatically call pfInitGfx() and will automatically create a new
pfState for your window. If you have your own window management and do not call
pfOpenWin(), then you should definitely call pfInitGfx() to initialize the window’s
graphics state for OpenGL Performer rendering. You will also need to call pfNewState()
to create a pfState for OpenGL Performer’s state management.
For X windows, OpenGL Performer maintains two windows and a graphics context. The
top level X window is placed on the screen and is the one that you should use in your
application for selecting X events. This top level window is very lightweight and has
minimal resources allocated to it. OpenGL Performer then maintains a separate X
window that is a child of the parent X window and is the one that is attached to the
graphics context. This allows you to select different framebuffer resources for the same
drawing area by just selecting a different graphics window and graphics context pair for
the parent X window. pfWindows directly support this functionality and this is discussed
in the next section, “Manipulating a pfWindow”. Finally, with OpenGL, you may choose
to draw to a different X Drawable than a window. X windows are created with the X
function XCreateWindow(). OpenGL graphics contexts are created with
glXCreateContext(). The parent X Window can be set with pfWinWSWindow(), the
graphics window or X Drawable is set with pfWinWSDrawable() and can be an X
window, pbuffer, or pixmap. The graphics context is set with pfWinGLCxt(). OpenGL
Performer defines the following window-system-independent types defined in
Table 13-3. If you create your own window but want to use pfQueryWin(), you must also
provide the framebuffer configuration information with pfWinFBConfig().
pfQueryWin() uses the internally stored visual.
Table 13-3

Window System Types

pfWS Type

X Type

pfWindow Set/Get Routine

pfWSWindow

X Window for IRIX and Linux
HWND for Microsoft Windows

pfWinWSWindow()
pfGetWinWSWindow()

pfWSDrawable

Drawable (window, pbuffer, pixmap)
for IRIX and Linux

pfWinWSDrawable()
pfGetWinWSDrawable()

HANDLE (to window or buffer) for
Microsoft Windows

460

007-1680-080

Manipulating a pfWindow

Table 13-3

Window System Types (continued)

pfWS Type

X Type

pfGLContext

OpenGL:
Linux

pfWindow Set/Get Routine

GLXContext for IRIX and pfWinGLCxt()
pfGetWinGLCxt()

HGLRC for Microsoft Windows
pfFBConfig

XVisualInfo* or GLXFBConfigSGIX*
for IRIX and Linux

pfWinFBConfig()
pfGetWinFBConfig()

int (pixel format ID) for Microsoft
Windows
pfWSConnection

Display* for IRIX and Linux only

pfGetCurWSConnection()

Manipulating a pfWindow
Windows are opened with pfOpenWin() and closed with pfCloseWin(). When a
window is closed, its graphics context is deleted. If you have multiple windows, you
select the window to draw to with pfSelectWin(). Multiple windows can be made more
efficient using share groups configured with pfWinShare() to share hardware resources.
Multiple windows can be made to have swapbuffers execute simultaneously through
window swap groups created with pfAttachWinSwapGroup(). There are also some
additional modes on pfWindows to control their behavior under various operations. This
section goes through the basics of these important features.
There are some modes you can set that can effect the general look and behavior of your
window and alternate configuration windows. These boolean modes can be individually
set and changed at any time with pfWinMode() and the tokens in Table 13-4.

007-1680-080

461

13: Windows

Table 13-4

pfWinMode() Tokens

PFWIN_ Token

Description

NOBORDER

Window will be without normal window system border

HAS_OVERLAY

Overlay alternate configuration window will be managed by the pfWindow.

pfOpenWin() will automatically create an overlay window if one has not
already been set.

pfWinIndex(win, PFWIN_OVERLAY_WIN) will also automatically
create and open an overlay window if one has not already been set.
HAS_STATS

Statistics alternate configuration window will be managed by the
pfWindow. pfOpenWin() will automatically create a statistics window if
one has not already been set.
pfWinIndex(win, PFWIN_OVERLAY_WIN) will also automatically
create and open a statistics window if one has not already been set and if the
current window cannot support statistics.

AUTO_RESIZE

The graphics window and active alternate configuration windows are
automatically resized to match the parent pfWinWSWindow(). This
mode is enabled by default.

ORIGIN_LL

The origin of the pfWindow, for placement purposes, will be the lower-left
corner. X and the Microsoft Windows GUI use the upper left corner as the
origin. This mode is enabled by default.

EXIT

The application will receive a DeleteWindow message upon selection of the
“Exit” from the window system menu on the window border.

Alternate Framebuffer Configuration Windows
OpenGL Performer supports multiple framebuffer configurations for the same drawing
area with alternate configuration windows. An OpenGL Performer alternate
configuration window has the same window parent (pfWinWSWindow()) but may have
a different drawable and graphics context. There are standard alternate configuration
windows for overlay and statistics windows that can be automatically created upon
demand.
An alternate configuration window is created as a full pfWindow and is an alternate
configuration window by virtue of being given to a base window in a pfList of alternate
configuration windows, or being directly assigned as one of the standard alternate

462

007-1680-080

Manipulating a pfWindow

configuration windows with either of pfWinOverlayWin() or pfWinStatsWin(). A
pfWindow may be an alternate configuration window of only one base window at a time;
alternate configuration windows may not be instanced between base windows. The
sharing of window attributes between alternate configuration windows, such as the
parent X window and GL objects (for OpenGL windows), must be set with pfWinShare()
on the base window and applied to the alternate configuration windows with
pfAttachWin(). You select the desired alternate configuration window to draw into with
pfWinIndex() and provide an index into your alternate configuration window list or one
of the standard indices (PFWIN_GFX_WIN, PFWIN_OVERLAY_WIN, or
PFWIN_STATS_WIN). PFWIN_GFX_WIN is the default window index and selects the
base window. If the alternate configuration window has not been opened, it will be
opened automatically upon being selected for rendering. Example 13-2 demonstrates
creating a pfWindow using the default overlay window. The graphics drawable and
graphics context of an alternate configuration window of a pfWindow can be closed with
pfCloseWinGL(). This can be called on the base window, in which case the active
alternate configuration window’s GL window and context will be closed, or it can be
called on the alternate configuration window pfWindow directly. The main parent
window will remain on the screen and a new alternate configuration window can be
applied to it or pfOpenWin() can be called to create a new graphics window and context.

Window Share Groups
Multiple windows on a screen will require duplicate processing and resources unless
they are set up as share groups. A pfWindow is attached to the group of another with
pfAttachWin(groupWin, attachee). The attributes to be shared are set with pfWinShare()
on any of the windows in the group. The full list of attributes are in the man page for
pfWindow (and similarly for pfPipeWindow) but most notably are
PFWIN_SHARE_GL_CXT for using the same graphics context across multiple windows,
PFWIN_SHARE_STATE for sharing full state information, and
PFWIN_SHARE_GL_OBJS for sharing display lists and textures across windows. In
particular, sharing GL objects is important if a display list (such as for fonts) are to be
created for one context and used in multiple contexts. When default alternate
configuration windows are automatically created (overlay and stats) they are configured
to share GL objects with the base window. A libpf pfPipeWindow example of window
share groups is in /usr/share/Performer/src/pguide/libpf/C/multiwin.c
for IRIX and Linux and in %PFROOT%/Src/pguide/libpf/C/multiwin.c for
Microsoft Windows.

007-1680-080

463

13: Windows

Synchronization of Buffer Swap for Multiple Windows
On IRIX systems, double-buffered pfWindows in window swap groups will have
simultaneous hardware execution of the buffer swap. There is a similar mechanism for
pfPipeWindows activated through pfChannel share groups sharing
PFCHAN_SWAPBUFFERS_HW that is discussed in Chapter 2, “Setting Up the Display
Environment.”
A window swap group is created by attaching windows with
pfAttachWinSwapGroup(groupWin, attachee). There is no global list maintained for
the swap group and their status so you cannot get back a list of windows in the group.
However, pfWinInSwapGroup() returns 1 if the specified window as been synchronized
to a swap group and 0 otherwise. This synchronization configuration will actually take
place upon a call to pfSelectWin() for the window. Windows of separate screens can be
attached but this also requires a BNC cable (of any Ohms) to be attached to the swap
ready connectors of the graphics pipelines. Detach from swap groups is not supported.
GLX barriers are used for multi-pipeline synchronization. If necessary, you can have a
pfWindow explicitly join a specific barrier group with pfWinSwapBarrier(). When
windows of multiple screens are attached, the video vertical retrace of those screens
should also be syncrhonized with genlock(7).

Communicating with the Window System
You can communicate with a local or remote window server by means of a window
system connection, a pfWSConnection (in X, also known as a Display connection). You
can use your pfWSConnection for selecting X events for your window, as is
demonstrated in Example 13-4.
libpr offers several utilities for creating a connection to a window server. A given
connection can communicate with any screen managed by that window server so usually
a process only needs one connection. A process should not share the connection of
another process, so you will need a connection per process. Typically, there is exactly one
window server on a machine but that is not required. libpr maintains a
pfWSConnection for the current process. By default, this connection obeys the setting of
the DISPLAY environment variable which can point to a window server on a local or a
remote machine. The current connection can be requested with
pfGetCurWSConnection() and can be set with pfSelectWSConnection(). Whenever
possible, use this connection to limit the total number of open connections.
pfOpenScreen() is a convenient mechanism for opening a connection with a specified

464

007-1680-080

More pfWindow Examples

default screen. pfOpenWSConnection() allows you to specify the exact name specifying
the desired target for the connection. Both pfOpenScreen() and
pfOpenWSConnection() allow you to specify if you would like the new connection to
automatically be made the current libpr pfWSConnection; this is recommended.

More pfWindow Examples
Example 13-2 demonstrates the creation of a window with a default overlay window.
Example 13-2

Using the Default Overlay Window

int main (void)
{
pfWindow *win, *over;
/* Initialize Performer */
pfInit();
pfInitState(NULL);
/* Initialize the window. */
win = pfNewWin(NULL);
pfWinOriginSize(win, 100, 100, 500, 500);
pfWinName(win, “OpenGL Performer”);
pfWinType(win, PFWIN_TYPE_X);
pfWinMode(win, PFWIN_HAS_OVERLAY, 1);
pfOpenWin(win);
/* First select and draw into the overlay window */
pfWinIndex(win, PFWIN_OVERLAY_WIN);
/* Select causes the index to be applied */
pfSelectWin(win);
...
/* Then select the main gfx window */
pfWinIndex(win, PFWIN_GFX_WIN);
pfSelectWin(win);
...
}

Example 13-3 demonstrates creating a custom overlay window and is taken from the
sample program
/usr/share/Performer/src/pguide/libpr/C/winfbconfig.c for IRIX and
Linux and %PFROOT%/Src/pguide/libpr/C/winfbconfig.c for Microsoft
Windows.

007-1680-080

465

13: Windows

Example 13-3

Creating a Custom Overlay Window

static int OverlayAttrs[] = {
PFFB_LEVEL, 1, /* Level 1 indicates overlay visual */
PFFB_BUFFER_SIZE, 8,
None,
};
int main (void)
{
pfWindow *win, *over;
/* Initialize Performer */
pfInit();
pfInitState(NULL);
/* Initialize the window. */
win = pfNewWin(NULL);
pfWinOriginSize(win, 100, 100, 500, 500);
pfWinName(win, “OpenGL Performer”);
pfWinType(win, PFWIN_TYPE_X);
pfWinMode(win, PFWIN_HAS_OVERLAY, 1);
over = pfNewWin(NULL);
pfWinName(over, “OpenGL Performer Overlay”);
pfWinType(over, PFWIN_TYPE_X | PFWIN_TYPE_OVERLAY);
/* See if we can get the desired overlay visual */
if (!(pfChooseWinFBConfig(over, OverlayAttrs)))
pfNotify(PFNFY_NOTICE, PFNFY_PRINT,
“pfChooseWinFBConfig failed for OVERLAY win”);
pfOpenWin(win);
/* First select and draw into the overlay window */
pfWinIndex(win, PFWIN_OVERLAY_WIN);
/* Select causes the index to be applied */
pfSelectWin(win);
...
/* Then select the main gfx window */
pfWinIndex(win, PFWIN_GFX_WIN);
pfSelectWin(win);
...
}

466

007-1680-080

More pfWindow Examples

Example 13-4 demonstrates the selection of X input events on a pfWindow. This example
is taken from /usr/share/Performer/src/pguide/libpr/C/hlcube.c for IRIX
and Linux and from %PFROOT%/Src/pguide/libpr/C/hlcube.c for Microsoft
Windows.. See the /usr/share/Performer/src/pguide/libpf/C/complex.c
sample program for a detailed example of using either standard or forked X input on
pfWindows for IRIX and Linux and sample program
%PFROOT%/Src/pguide/libpf/C/complex.c for Microsoft Windows.
Example 13-4

pfWindows and X Input

pfWSConnection Dsp;
void main (void)
{
pfWindow *win;
pfWSWindow xwin;
/* Initialize Performer */
pfInit();
pfInitState(NULL);
/* Initialize the window. */
win = pfNewWin(NULL);
pfWinOriginSize(win, 100, 100, 500, 500);
pfWinName(win, “OpenGL Performer”);
pfWinType(win, PFWIN_TYPE_X);
pfOpenWin(win);
...
/* set up X input event handling on pfWindow */
Dsp = pfGetCurWSConnection();
xwin = pfGetWinWSWindow(win);
XSelectInput(Dsp, xwin, KeyPressMask );
XMapWindow(Dsp, xwin);
XSync(Dsp,FALSE);
...
do_events(win);
}
static void
do_events(pfWindow *win)
{
while (1) {
while (XPending(dsp))
{
XEvent event;
XNextEvent(Dsp, &event);

007-1680-080

467

13: Windows

switch (event.type)
{
case KeyPress:
....
}
}
}

468

007-1680-080

Chapter 14

14. pfPipeWindows and pfPipeVideoChannels

This chapter describes the additional windowing and video-channel based functionality
provided by pfPipeWindows and pfPipeVideoChannels. These libpf objects, based on
their libpr counterparts, pfWindows and pfVideoChannels, provide automatic
configuration, multiprocessing, and extended functionality by being hooked together
with pfChannels.

Using pfPipeWindows
OpenGL Performer can automatically create and open a full screen window with a
default configuration for your pfPipe. At the other extreme, you can create and configure
your own windows and set them on a pfPipe. There is a single interface for creating,
configuring, and managing the windows. The pfPipeWindow is the mechanism by
which a pfPipe knows about and keeps track of the windows to which it is to render, the
size of the render area, and the framebuffer configuration. pfPipes and pfChannels need
this information for proper viewport and frustum management and for using rendering
features like antialiasing, transparency for fade LOD, and layers for decal geometry that
are all affected by framebuffer configuration.
In the simplest case, OpenGL Performer automatically creates a pfPipeWindow for the
application and automatically open a full screen window upon the first call to pfFrame().
This trivial case is demonstrated in Example 14-1.

Creating, Configuring and Opening pfPipeWindow
In most cases, there are some window parameters, such as size and origin, that you will
want to set. You may also have custom graphics state that you need to set to fully
initialize your rendering window. This section describes the basics for setting up
windows through the pfPipeWindow mechanism.

007-1680-080

469

14: pfPipeWindows and pfPipeVideoChannels

A pfPipeWindow can be created for a pfPipe using pfNewPWin(pipe). If you create a
pfPipeWindow, then you are responsible for explicitly opening it. The call to
pfOpenPWin(pwin) from the application process causes the next call to pfFrame() to
trigger the opening of the pfPipeWindow in the draw process. A pfPipeWindow created
in the application will be a rubber-band window of undefined size for the user to stretch
out. This is in contrast to the full screen window that OpenGL Performer creates on your
behalf in the fully automatic case. To easily get a full screen window, you can use the
pfPWinFullScreen() function. pfPWinOriginSize() can be used to set a fixed position
and size for the window. The code in Example 14-1, placed in the application process,
creates and opens a window in the lower-left corner of the screen of size 500 pixels on
each side.
Example 14-1

Creating a pfPipeWindow

main()
{
pfPipe *pipe;
pfPipeWindow *pwin;
pfInit();
....
pfConfig();
/* Create pfPipeWindow for pfPipe 0 */
pipe = pfGetPipe(0;
pwin = pfNewPWin(pipe);
/* Set the origin and size of the pfPipeWindow */
pfPWinOriginSize(pwin, 0, 0, 500, 500);
/* Tell OpenGL Performer that the pfPipeWindow is ready to
* be opened
*/
pfOpenPWin(pwin);
/* trigger the opening of the pfPipeWindow
* in the draw process
*/
pfFrame();
...
while(!SimDone())
{ ... }
}

The pfPipeWindow is always physically opened in the draw process when processing
the application frame that requested the window to be opened. When both the cull and
draw processes are running as separate processes, there might be a 2-frame delay (two
additional calls to pfFrame()) for the window do actually be opened. Additionally, if the
draw is running as a separate process, the window will not be opened right after pfFrame

470

007-1680-080

Using pfPipeWindows

but some time in that following frame. If the pfPhase is such that the application process
is allowed to spin ahead while the draw process does expensive initialization (anything
but PFPHASE_FREE_RUN), the application process may execute many pfFrame() calls
before the window is physically opened in the draw process. If in the application process
you need to check on a result from opening the window, such as framebuffer
configuration, you will want to do something that is in effect equivalent to the following:
while (!pfIsPWinOpen(pwin)) pfFrame();

pfPipeWindows are actually built upon libpr pfWindows, but have added support for
handling the multiprocessed environment of libpf applications and fit into the libpf
display hierarchy of pfPipes, pfPipeWindows, and pfChannels. Additionally,
pfPipeWindows support the multiprocessing environment of libpf by having a
separate copy of each pfPipeWindow in each pipeline process. All of the “windowness”
of pfPipeWindows really comes from the fact that there is a pfWindow internal to the
pfPipeWindow. Many of the basic support routines, such pfPWinFullScreen() and
pfWinFullScreen(), have very similar functionality for pfWindows and pfPipeWindows.
However, there are situations where pfPipeWindows are able to provide the same
functionality in a much more efficient manner. Management of dynamic window origin
and size is one case where pfPipeWindows have a real advantage over pfWindows.
pfPipeWindows are able to take advantage of the multiprocessed libpf environment to
always be able to return an accurate window size and origin relative to the window
parent. A process separate from the rendering process is notified by the window system
of changes in the pfPipeWindow’s size in an efficient manner without impacting the
window system or the rendering process. This can be forced off for the real-time static
displays of a deployed visual simulation system by making the pfPipeWindow of type
PFPWIN_TYPE_NOXEVENTS, which prevents OpenGL Performer from tracking the
window. Further details regarding basic window creation and configuration are
discussed with pfWindows in Chapter 13, “Windows.”
Note: pfPWin*() routines expect a pfPipeWindow and the pfWin*() routines a
pfWindow. These routines are not interchangeable: pfWindow routines cannot accept
pfPipeWindows nor the reverse. The PFWIN_* tokens can be used with the
pfPipeWindow routines.
Windows have some intrinsic type attributes that must be set before the window is
opened. The selection of the screen of a window is determined by the pfPipe that it is
opened on, set for both the pfWindow and its pfPipe with the call pfPWinScreen(), or
else set by the value of the DISPLAY variable when the window is finally opened. The
window system configuration of the window must also be set before the window is

007-1680-080

471

14: pfPipeWindows and pfPipeVideoChannels

opened. Windows under OpenGL operation on IRIX and Linux systems will always be
X windows. On Microsoft Windows, the windows will be HWNDs.
Note: Micrsoft Wndows virtualizes the concept of a screen as well as all available
hardware. Hence, on such systems, you cannot specify the screen connection in a manner
similar to using the DISPLAY environment variable.
An open window must be closed for its type to be changed. The window type argument
is actually a bitmask and the type of a pfPipeWindow can include the attributes listed in
Table 14-1.
Table 14-1

pfPWinType Tokens

PFPWIN_TYPE_*
Bitmask Token
Type Attributes

The default. Rendering will be done to an X window. Ignored by OpenGL as all
OpenGL rendering is done to X windows.

STATS

The window’s normal drawing configuration supports graphics statistics. This
affects framebuffer configuration and fill statistics.

The pfPipeWindow automatically attaches to the first pfPipeWindow of the
parent pipe with pfAttachPWin().

PBUFFER

The window drawable is a pbuffer (not visible on the screen).

NOXEVENTS

Window size and position tracking is not done.

UNMANAGED No automatic window management operations other than select for rendering
happens. Window is not auto-sized or tracked. Swapbuffers will not
automatically be done.

pfPipeWindows have a target default framebuffer configuration. The ability to meet this
target depends on the current graphics hardware configuration, as well as their type. The
following parameters are part of the target default configuration and are listed in their
order of priority. If the goal framebuffer configuration cannot be created on the current
graphics hardware configuration, lower priority parameters are downgraded as
specified in the following:
1.

double buffered

2. RGB mode with 8 bits per color component (4 if 8 cannot be supported)

472

007-1680-080

Using pfPipeWindows

3. z-buffer with depth of 23 or 24 bits, as available
4. one-bit stencil buffer (window type PFWIN_TYPE_STATS still requires four bits of
stencil)
5. multisample buffer of 8, 4, or 0 samples as available
6. four-bit stencil buffer if still available after the above is satisfied
pfPipeWindows have OpenGL Performer libpr rendering state automatically
initialized when they are opened. Additionally the following graphics state is
automatically initialized when a window is opened or upon any call to pfInitGfx() for
an open window:
1.

RGB mode is enabled.

2. z-buffer is enabled and a z range is set.
3. Viewport clipping is enabled.
4. subpixel vertex accuracy is enabled.
5. The viewing matrix is initialized to a two-dimension, one-to--one mapping from eye
coordinates to window coordinates.
6. The model matrix is initialized to the identity matrix and made the current GL
matrix.
7. Backface removal is enabled.
8. Smooth shading is enabled.
9. If the current graphics hardware platform supports multisampling, multisampled
antialiasing will be enabled with pfAntialias(PFAA_ON).
10. A default modulating texture environment is created.
11. A default lighting model is created.
Custom framebuffer configuration for a pfPipeWindow can be specified with
pfPWinFBConfigAttrs(), pfPWinFBConfig(), and pfChoosePWinFBConfig(). These
routines have identical functionality as each of the corresponding pfWindow routines.
However, the function pfChoosePWinFBConfig() has the constraint that it be called in
the draw process because it creates and stores internal data from the window server that
must be kept local to the process in which it is called. Table 14-2 lists the different
pfPipeWindow routines and describes multiprocessing constraints.

007-1680-080

473

14: pfPipeWindows and pfPipeVideoChannels

The flexibility in changing the framebuffer configuration of OpenGL/X windows or
OpenGL/HWNDs is complex. The main window can remain in place but structures
under it must be switched or replaced. If multiple framebuffer configurations are likely
to be desired, multiple graphics contexts can be created for the window using
pfWindows. pfPipeWindows and pfWindows both allow you to have a list of alternate
pfWindows that render to exactly the same screen area but may have different
framebuffer configuration. You then select the current configuration for a pfPipeWindow
with pfPWinIndex(). There are two kinds of common alternate configuration windows
that can be created automatically for you: overlay windows created in the overlay planes
and windows to support hardware fill statistics (discussed in Chapter 20, “Statistics”).
You can use pfPWinMode() to indicate that you would like these windows created for
you automatically. Special tokens to pfPWinIndex() are used to select these common
special alternate configuration windows—PFWIN_GFX_WIN,
PFWIN_OVERLAY_WIN and PFWIN_STATS_WIN—where PFWIN_GFX_WIN selects
the normal default drawing window. Note that only a pfWindow, never a
pfPipeWindow, can be an alternate configuration window. The source code in
Example 14-2 is taken from
/usr/share/Performer/src/pguide/libpf/C/pipewin.c for IRIX and Linux
and from %PFROOT%/Src/pguide/libpf/C/pipewin.c for Microsoft Windows. It
demonstrates the automatic creation and selection of overlay and statistics windows for
a pfPipeWindow. This also shows usage of pfChannels and interaction between
pfPipeWindows and pfChannels discussed in the section “Creating and Configuring a
pfChannel” in Chapter 2.
Example 14-2

pfPipeWindow With Alternate Configuration Windows for Statistics

main()
{
pfPipe *pipe;
pfPipeWindow *pwin;
pfInit();
....
pfConfig();
/* Create pfPipeWindow for pfPipe 0 */
pipe = pfGetPipe(0);
pwin = pfNewPWin(pipe);
/* request automatic default overlay and stats windows */
pfPWinMode(pwin, PFWIN_HAS_OVERLAY, PF_ON);
pfPWinMode(pwin, PFWIN_HAS_STATS, PF_ON);
/* Open the main graphics window */
pfOpenPWin(pwin);
pfFrame();

474

007-1680-080

Using pfPipeWindows

while(!SimDone())
{
...
if (Shared->winSel == PFWIN_STATS_WIN))
{
/* select statistics window and enable fill stats */
pfPWinIndex(Shared->pw, PFWIN_STATS_WIN);
pfFStatsClass(fstats,
PFSTATSHW_ENGFXPIPE_FILL, PFSTATS_ON);
pfEnableStatsHw(PFSTATSHW_ENGFXPIPE_FILL);
}
else
{
/* we are not doing statistics so turn them off */
pfFStatsClass(fstats,
PFSTATSHW_ENGFXPIPE_FILL, PFSTATS_OFF);
pfDisableStatsHw(PFSTATSHW_ENGFXPIPE_FILL);
pfPWinIndex(Shared->pw, Shared->winSel);
...
}
}
/* Channel draw process drawing function */
void DrawFunc(void pfChannel *chan)
{
pfPipeWindow *pwin;
pwin = pfGetChanPWin(chan);
if (pfGetPWinIndex(pwin) == PFWIN_OVERLAY_WIN)
{
/* Draw overlay image */
DrawOverlay();
/* Put back the normal drawing window */
pfPWinIndex(pwin, PFWIN_GFX_WIN);
/* Indicate that we will now draw to the window */
pfSelectPWin(pwin);
}
/* call the main OpenGL Performer drawing function */
pfDraw();
}

Notice that in Example 14-2, although the pfPipeWindow is double buffered, the front
and back color buffers are never explicitly swapped. For pfPipeWindows, this operation
is done automatically after all channels on the parent pfPipe have completed their

007-1680-080

475

14: pfPipeWindows and pfPipeVideoChannels

drawing for the given frame. The color buffers of a pfPipeWindow may be swapped
explicitly with pfSwapWinBuffers. This call may be placed in a user-swap function call
back placed on the pfPipe with pfPipeSwapFunc() to replace the pfPipe normal swap
behavior. The swap callback will be called in the draw process at the end of the frame
after all pfChannels in all pfWindows have been drawn for the pfPipe. The function is
called for all of the pfPipeWindows on the pfPipe. This is additionally useful for doing
end-of-frame rendering or readbacks from the framebuffer.
You may need to set additional window and graphics state to complete the initialization
of your pfPipeWindow. Calling pfOpenPWin() from the application process does not
give you the opportunity to do this. However, with pfPWinConfigFunc(), you can
supply a window configuration callback function that will enable you to open and
initialize your pfPipeWindow in the draw process. A call to pfConfigPWin() triggers one
call of the window configuration callback in the draw process upon the next call to
pfFrame(). pfConfigPWin() can be called at any time to trigger the calling of the current
window configuration function in the draw process. Example 14-3 demonstrates
initializing a pfPipeWindow from a draw process window configuration callback. It
creates a global light to serve as the Sun in the window configuration callback. (see the
/usr/share/Performer/src/pguide/libpf/C/complex.c example for IRIX
and Linux and %PFROOT%/Src/pguide/libpf/C/complex.c for Microsoft
Windows).
Example 14-3

Custom Initialization of pfPipeWindow State

main()
{
pfPipe *pipe;
pfPipeWindow *pwin;
pfInit();
....
pfConfig();
/* Create pfPipeWindow for pfPipe 0 */
pipe = pfGetPipe(0);
pwin = pfNewPWin(pipe);
/* Set the configuration function for the pfPipeWindow */
pfPWinConfigFunc(pwin, OpenPipeWindow);
/* Indicate that OpenPipeWindow should be called in the
* draw process.
*/
pfConfigPWin(pwin);
/* trigger OpenPipeWindow to be called in the draw process */

476

007-1680-080

Using pfPipeWindows

pfFrame();
while(!SimDone())
{ ... }
}
/* Initialize graphics state in the draw process */
void
OpenPipeWindow(pfPipeWindow *pw)
{
/* Set some configuration stuff */
pfPWinOriginSize(pw, 0, 0, 500, 500);
/* Open the window - will give us initialized libpr and
graphics state
*/
pfOpenPWin(pw);

/* create a global light in the “south-west” (QIII) */
Sun = pfNewLight(NULL);
pfLightPos(Sun, -0.3f, -0.3f, 1.0f, 0.0f);
}

In Example 14-3 the functions pfPWinOriginSize() and pfOpenPWin() are now called
in the draw process, as opposed to the application process as in Example 14-1. In general,
configuring or editing any libpf object must be done in the application process.
pfPipeWindows must be created in the application process. However, pfPipeWindows
may be configured, edited, opened and closed in the pfPWinConfigFunc() configuration
callback which will be called in the draw process. Window operations are best done in a
configuration callback, though they can also be done in the drawing callback for a
pfChannel on the window. Any function which aspires to directly affect the graphics
context must be called in the drawing process. Table 14-2 shows which processes
(application or draw via a configuration function) that pfPipeWindow calls can be made
from and further detail about these functions can be found in the discussion of
pfWindows in Chapter 13, “Windows.”
Table 14-2

007-1680-080

Processes From Which to Call Main pfPipeWindow Functions

pfPipeWindow Function

Application Process

Draw Process

pfNewPWin()

Yes

pfPWinMode()

Yes

pfPWinIndex()

Yes

477

14: pfPipeWindows and pfPipeVideoChannels

Table 14-2

Processes From Which to Call Main pfPipeWindow Functions (continued)

pfPipeWindow Function

Application Process

Draw Process

pfPWinConfigFunc()

Yes

pfOpenPWin()
pfClosePWin()
pfClosePWinGL()

Yes

pfPWinOriginSize()
pfPWinFullScreen()

Yes

pfGetPWinCurOriginSize()

Yes

Yes.

pfPWinFBConfigAttrs()

Yes

Yes.

pfChoosePWinFBConfig()

Yes.

pfPWinFBConfig()

Yes, but the pfFBConfig* must be
valid for access in the draw process.

Yes.

pfPWinType()
pfPWinScreen()
pfPWinShare(), pfAttachWin()

Yes (before opened)

Yes (before
opened).

pfPWinWSWindow()

Yes

Yes.

pfPWinGLCxt()

Yes, but the context must be created
in the draw process.

Yes.

pfQueryWin()
pfMQueryWin()

Yes.

pfAddPWinPVChan()

Yes

Yes.

pfAttachPWinSwapGroup()

Yes

OpenGL Performer provides GL-independent framebuffer configuration utilities. In
most cases, pfPWinFBConfigAttrs(pwin, attrs) can be used to select a framebuffer
configuration for your pfPipeWindow based on the array of attribute tokens attrs. If attrs
is NULL, the default framebuffer configuration will be selected. If attrs is not NULL, the
rules for default values follow the rules for configuring windows in OpenGL and X,
which are different from values in the OpenGL Performer default window configuration.
Such window framebuffer configuration should be done in the draw process in a
window configuration callback function before the call to pfOpenPWin(). Window

478

007-1680-080

Using pfPipeWindows

framebuffer configuration for pfPipeWindows is identical to that of pfWindows and is
discussed in more detail in Chapter 13, “Windows,” but the following is a simple
example of the specification of framebuffer configuration taken from the sample source
code example program
/usr/share/Performer/src/pguide/libpf/C/pipewin.c for IRIX and Linux
and from %PFROOT%/Src/pguide/libpf/C/pipewin.c for Microsoft Windows:
Example 14-4

Configuration of a pfPipeWindow Framebuffer

static int FBAttrs[] = {
PFFB_RGBA,
PFFB_DOUBLEBUFFER,
PFFB_DEPTH_SIZE, 24,
PFFB_RED_SIZE, 8,
PFFB_SAMPLES, 8,
PFFB_STENCIL_SIZE, 1,
NULL,
};
main()
{
pfPipe *pipe;
pfPipeWindow *pwin;
pfInit();
....
pfConfig();
/* Create pfPipeWindow for pfPipe 0 */
pipe = pfGetPipe(0);
pwin = pfNewPWin(pipe);
/* Set the framebuffer configuration */
pfPWinFBConfigAttrs(Shared->pw, FBAttrs);
/* Indicate that the window is ready to open */
pfOpenPWin(pwin);
/* trigger the opening of the window in the draw */
pfFrame();
...
}

If you want to do all of your own window creation and management you can do so and
just give OpenGL Performer the handles to your windows with the
pfPWinWSDrawable() function; you may also provide a parent X window (or HWND
on Microsoft Windows) with the pfPWinWSWindow() function. pfOpenPWin() makes
use of any windows that have already been provided. More details regarding the

007-1680-080

479

14: pfPipeWindows and pfPipeVideoChannels

creation and configuration of pfPipeWindows and pfWindows are discussed in
Chapter 13, “Windows.”

pfPipeWindows in Action
pfPipeWindows allow for a reasonable amount of flexibility in the running application.
pfPipeWindows can be re-ordered on their parent pfPipe to control the order that they
are drawn in with the command pfMovePWin(pipe, index, pwin). pfPipeWindows can be
dynamically opened and closed in the application or draw processes with
pfOpenPWin() and pfClosePWin(). Additionally, pfConfigPWin() can be re-issued at
any time from the application process to call the current window configuration function
to dynamically open, close, and reconfigure pfPipeWindows.
The following example is taken from the distributed source code example file
/usr/share/Performer/src/pguide/libpf/C/pipewin.c for IRIX and Linux
and from %PFROOT%/Src/pguide/libpf/C/pipewin.c for Microsoft Windows. It
demonstrates the dynamic closing of a window from the application process in the
simulation loop and the reuse of pfConfigPWin() to reopen the window.
Example 14-5

Opening and Closing a pfPipeWindow

main()
{
...
/* main simulation loop */
while (!Shared->exitFlag)
{
/* wait until next frame boundary */
pfSync();
pfFrame();
/* Set view parameters for next frame */
UpdateView();
pfChanView(chan, Shared->view.xyz, Shared->view.hpr);
/* Close pfPipeWindow */
if (Shared->closeWin == 1)
{
pfClosePWin(Shared->pw);
ct = pfGetTime();
Shared->closeWin = 2;
}
/* then wait two seconds and reconfig window */

480

007-1680-080

Using pfPipeWindows

}

else if ((Shared->closeWin == 2) &&
(pfGetTime() - ct > 2.0f))
{
pfConfigPWin(Shared->pw);
Shared->closeWin = 3;
pfNotify(PFNFY_NOTICE, PFNFY_PRINT, “OPEN”);
}
...

}

Motif

You may want your windows to reside within a larger Motif interface and window
hierarchy. OpenGL Performer supports this and allows you to run the Motif main loop
in a separate process so that you can maintain control of your simulation loop. The Motif
interface is created in its own process and Motif event handlers and callbacks will be
executed in that process. The Motif callbacks set flags in shared memory to communicate
with the main application. Part of this communication is the sharing of X windows
between OpenGL Performer and Motif. The example program
/usr/share/Performer/src/pguide/libpf/C/motif.c for IRIX and Linux and
program %PFROOT%/Src/pguide/libpf/C/motif.c for Microsoft Windows
demonstrate the basic elements of this integrated OpenGL Performer-Motif hook-up.
Multiple pfPipeWindows and Multiple pfPipes

The use of multiple windows on a single graphics pipe can add overhead. The sharing of
the graphics context between windows consumes almost all of this overhead. To simply
share a single graphics context across windows of two pfPipe objects, include
PFPWIN_TYPE_SHARE in the pfPWinType() call. The sharing of pfPipeWindows and
attributes can be completely controlled by setting up the sharing manually to create
pfPipeWindow share groups with pfPWinAttach(groupPWin, attachee) and
pfPWinShare() as is done with pfWindows, discussed in Chapter 13, “Windows.”
pfPipeWindows can have pfWindows in their share group if a pfPipeWindow is the main
group window.
Multiple windows, particularly those on separate graphics pipelines, that are intended
to produce results that can be seen as a single image, such as projected side by side on a
large screen or to video outputs used for a stereo display, must have their video vertical
retraces synchronized with genlock(7) and their double buffering synchronized. This
is necessary for both image quality and performance reasons as the last window to finish
operation can hold up all of the rendering processes. Window double-buffering

007-1680-080

481

14: pfPipeWindows and pfPipeVideoChannels

synchronization can be done through pfChannel share groups with the
PFCHAN_SWAPBUFFERS_HW token, as discussed in Chapter 2, “Setting Up the
Display Environment,” or explicitly by attaching the windows to form window swap
groups with pfAttachPWinSwapGroup(groupPWin, attachee) as discussed in Chapter 13,
“Windows.” pfPipeWindows can have pfWindows in their swap groups if a
pfPipeWindow is the main group window. The sample program in file
/usr/share/Performer/src/pguide/libpf/C/multipipe.c for IRIX and
Linux and in file %PFROOT%/Src/pguide/libpf/C/multipipe.c for Microsoft
Windows demonstrates multipipe synchronization.

Controlling Video Displays
You use pfPipeVideoChannel to direct the output of pfChannels to specified video
displays, as shown in Figure 14-1. OpenGL Performer uses the XSGIvc(3), which is a
Silicon Graphics extension of the X library, for video channel management.
pfPipe

pfChannel
pfPipeVideoChannel
(channels 0, 1, 2)

pfPipeWindow

Figure 14-1

Directing Video Output

pfPipeVideoChannels are based on pfVideoChannels; however, pfPipeVideoChannels
are maintained by libpf and are used by libpf to render the output of pfChannels
within a pfPipeWindow. The pfVideoChannel API is duplicated for

482

007-1680-080

Controlling Video Displays

pfPipeVideoChannels. pfPVChan*() methods operate on a pfPipeVideoChannel and
pfVChan*() methods operate on a pfVideoChannel.
There are several sample programs to help understand the use of pfVideoChannels and
pfPipeVideoChannels. The libpr program
/usr/share/Performer/src/pguide/libpr/C/queryvchan.c for IRIX and
Linux and the program %PFROOT%/Src/pguide/libpr/C/queryvchan.c for
Microsoft Windows show how to query video channel attributes through OpenGL
Performer and how to query additional attributes directly through the XSGIvc interface.
The program /usr/share/Performer/src/pguide/libpr/C/vchan.c for IRIX
and Linux and the program %PFROOT%/Src/pguide/libpr/C/vchan.c for
Microsoft Windows show basic video channel creation. On an InfiniteReality, this
example does resizing and translation of the video channel output area. The libpf
program /usr/share/Performer/src/pguide/libpf/C/pvchan.c for IRIX and
Linux and the program %PFROOT%/Src/pguide/libpf/C/pvchan.c for Microsoft
Windows show basic pfPipeVideoChannel creation and hookup.

Creating a pfPipeVideoChannel
You create a pfPipeVideoChannel from a pfPipe using pfNewPVChan() and providing
the parent pfPipe. The pfPipeVideoChannel is then attached to a pfPipeWindow with
pfAddPWinPVChan(), which returns an index into the pfPipeWindow’s video channel
list. Additionally, if the pfPipeVideoChannel has not already been assigned a hardware
video channel with pfPVChanId(), the next active video channel will be assigned.
pfPipe *p = pfGetPipe(0);
pfPipeVideoChannel *pvc = pfNewPVChan(p);
pfPVChanId(pvc,0);
pvcIndex = pfPWinAddPVChan(pwin, pvc);

To find out if a pfPipeVideoChannel with its current video channel assignment is active
for displaying output, call pfIsPVChanActive(), which returns 0 if the video channel
assignment is not fully defined or if the channel is not active and returns 1 otherwise. The
assignment of a hardware video channel is not complete until the screen of the pfPipe is
known and so might not be done immediately if the pfPipe screen has not been set with
pfPipeScreen() or if the pfPipeWindow is not open. Even if you have explicitly assigned
a hardware video channel ID, it is only meaningful relative to a known screen. The
hardware video channel assignment can be changed at any time with pfPVChanId(). It
is an error to have multiple pfPipeWindows and pfPipeVideoChannels attempt to
manage the same hardware video channel. The index returned by pfAddPWinPVChan()
is then used for a pfChannel to reference this pfPipeVideoChannel. pfPipeWindows

007-1680-080

483

14: pfPipeWindows and pfPipeVideoChannels

always have an initial pfPipeVideoChannel already attached whose default video
channel will be the first active video channel on the screen of the pfPipe. So, only add
pfPipeVideoChannels if you actually need additional video channels.

Multiple pfPipeVideoChannels in a pfPipeWindow
The only video channels a pfPipeVideoChannel can manage are those of the screen of the
pfPipe. Use pfGetNumScreenPVChans() to find out how many active video channels
are on a given screen.
To add additional pfPipeVideoChannels to a single pfPipeWindow, use
pfPWinAddPVChan(). The routine returns an index number associated with the
pfPipeWindow, or -1 if an error occurs. If the pfPipeVideoChannel does not already have
an assigned hardware video channel, the next active video channel relative to the video
channels already attached to the pfPipeWindow will be assigned.
There are list routines to manage the list of pfPipeVideoChannels on a pfPipeWindow.
Use pfGetPWinNumPVChans() to find out how many pfPipeVideoChannels are being
managed by the pfPipeWindow. pfPWinRemovePVChan() and
pfPWinRemovePVChanIndex() disassociate a pfPipeVideoChannel from a
pfPipeWindow, using either the pfPipeWindow object or an index number to specify the
pfPipeWindow.
There are additional utilities to find pfPipeVideoChannels on pfPipeWindows. You can
obtain the index value of a pfPipeVideoChannel for a pfPipeWindow using
pfGetPWinPVChanIndex(), which returns the index of a pfPipeVideoChannel, or -1 of
the pfPipeVideoChannel is not registered with the pfPipeWindow. Use
pfGetPWinPVChanId() to find a pfPipeVideoChannel on a pfPipeWindow with the
specified hardware Id. NULL will be returned if no such pfPipeVideoChannel exists on
the specified pfPipeWindow.

Configuring a pfPipeVideoChannel
pfPipeVideoChannels are bound to their hardware video channels in a lazy fashion as
needed by configuration requests. Basic queries of a video channel do not require any
explicit binding. Changing a video channel’s properties does require explicit binding.
pfPipeWindows manage this process. However, explicit binding and unbinding might
be necessary if changes are made to video channels directly through the XSGIvc API and
not through the pfPipeVideoChannel API. This is quite reasonable since

484

007-1680-080

Controlling Video Displays

pfPipeVideoChannels do not duplicate all of XSGIvc. A pfPipeVideoChannel is bound to
its hardware video channel with pfBindPVChan(). All of the pfPipeVideoChannels
associated with a pfPipeWindow can be bound in one step with pfBindPWinPVChans()
and unbound with pfUnbindPWinPVChans(). You can get the XSGIvc handle from a
pfPipeVideoChannel to do your own configuration or extended queries with
pfGetPVChanInfo().

Use pfPipeVideoChannels to Control Frame Rate
There are two mechanisms by which pfPipeVideoChannels can help you maintain
constant frame rate in your application. Dynamic Video Resolution (DVR) addresses
reducing load caused by filling to many pixels. Pan/Zoom video scan-out to sample
different parts of the frame buffer, under resize, to be selected asynchronously to a draw
process. Both of these mechanisms make use of editing the size and or origin of the
output area of a pfPipeVideoChannel, supported by InfiniteReality graphics platforms.
Using pfPVChanOutputSize() and pfPVChanOutputAreaScale() changes the output
area size of the bound video channel. pfPVChanOutputOrigin() changes the origin of
the output area. The pfPipeVideoChannel(3) references more routines to manage
video channel origin and size.
On InfiniteReality graphics platforms, you can use pfPipeVideoChannels to dynamically
adjust the size of the output area of the video channel. The output area is then
automatically zoomed up to full video channel size by the InfiniteReality hardware using
bilinear filtering. This operation has no added performance cost or latency. This feature
can be used to allow pfChannels to reduce their viewport size to the reduced video
channel output area. Reducing the number of pixels drawn reduces the fill load for the
pfPipe and can be used as a load management technique for maintaining constant frame
rates. pfPipeVideoChannels support manual resizing, allowing you to implement your
own load management, or automatically resizes the output area and the pfChannel
viewports. You can enable and select a resizing mode with pfPVChanDVRMode() and
providing PFPVC_DVR_MANUAL or PFPVC_DVR_AUTO. The default value is
PFPVC_DVR_OFF. For more information, see “Level-of-Detail Management” in
Chapter 5.
pfPipeVideoChannels also support asynchronous editing of their size and origin. This
can be used in an asynchronous process, or in an application process that is reliably
running at frame rate, to edit the origin and size of the video channel. These changes will
affect the following video field if the pfPVChanMode() for PFVCHAN_AUTO_APPLY
is set to 1 (default is 0) and the PFVCHAN_SYNC mode is set to

007-1680-080

485

14: pfPipeWindows and pfPipeVideoChannels

PFVCHAN_SYNC_FIELD (default is PFVCHAN_SYNC_FRAME which selects apply
upon swapbuffers).
Real-time changes to pfPipeVideoChannel origin or size should be done between
pfSync() and pfFrame() to affect the next draw frame or video field.

486

007-1680-080

Chapter 15

15. Managing Nongraphic System Tasks

This chapter describes objects that manage nongraphic tasks, including the following:
•

Queues

•

Clocks

•

Memory allocation

•

Asynchronous I/O

•

Error handling and notification

•

File search paths

Handling Queues
A pfQueue object is a queue of elements, which are all the same type and size; the default
size is the size of a void pointer. A pfQueue object actually consists of three interrelated
queues, as shown in Figure 15-1.
Input
buffer
Sorted
list
Output
buffer

pfQueue object

Figure 15-1

007-1680-080

pfQueue Object

487

15: Managing Nongraphic System Tasks

•

Input buffer—where processes dump values to be added to the pfQueue object

•

Output buffer—values at one end of the queue that processes may remove from the
output buffer pfQueue object

•

Sorted list—sorted values that processes may not remove from the pfQueue object

Note: In nonsorting mode, there is only the input buffer; values in the output buffer and
the sorted list are transferred into the input buffer.
Values in the input buffer are not sorted and are not part of the sorted list. Values in the
sorted list and the output buffer are sorted (when the pfQueue object is in sort mode)
according to a user-defined sorting function. Sorted values of highest priority are
automatically moved from the sorted list to the output buffer whenever the pfQueue
object is sorted. Priority is defined by the sorting function, for example, if a pfQueue
object contains pointers to tiles of texture, the sorting function might sort according to the
proximity of the viewer and the tile: the closer the tile is to the viewer, the higher its
priority, and the more likely the pointer to the tile will be in the output buffer. Processes
do not have access to values in the sorted list; only to those values in the output buffer.

Multiprocessing
Because there are separate input and output buffers, multiple processes can add or
retrieve elements, but only one process can actually insert elements into the input buffer
and one process retrieve elements from the output buffer at one time. The process adding
elements to the input buffer can be different from the process removing elements from
the output buffer.

Queue Contents
The contents of the pfQueue object can be any fixed-size object; for example, pfQueues
often contain pointers to OpenGL Performer objects. You might use a pfQueue object, for
example, to organize tiles of texture according to the direction the viewer is looking and
the proximity of the viewer to the tiles. Because you declare the size and type of objects
in the pfQueue in the constructor, you cannot change the type or size of its elements after
its creation.

488

007-1680-080

Handling Queues

Adding or Retrieving Elements
You can insert elements into the input buffer or remove them from the output buffer
using the following methods, respectively:
•

pfQueue::insert()

•

pfQueue::remove()

These methods can be used by multiple processes asynchronously without collision.

Warning: Do not insert NULL elements into the queue.
The pfQueue object is resized dynamically when the number of elements inserted into
the queue exceeds its declared size; the size is doubled automatically. Doubling the size
prevents repeated, incremental, costly resizing of the queue.
Tip: Doubling the size of the queue can cause excessive memory allocation. It is
important therefore to accurately declare the size of the queue.
You can set the size of the queue in the constructor of the pfQueue object or afterwards
by using pfQueue::setArrayLen(). pfQueue::getNum() returns the number of elements
in the queue.
Retrieving Elements from the Queue

It is possible for you to do the following:
1.

Create a thread to retrieve elements from the output buffer.

2. Use the pfQueue::remove() method to retrieve the element.
3. Delete the thread.
It is much easier, however, to use the pfQueue::addServiceProc() method to perform all
of those tasks. This method does the following:

007-1680-080

•

Creates a thread.

•

Returns the thread ID.

489

15: Managing Nongraphic System Tasks

•

Invokes the developer-supplied function in the argument of the function.

•

Deletes the thread.

The developer-supplied function must take as its argument an element from the output
buffer and process it. For example, if the queue contains pointers to tiles of texture, the
function might download a tile from disk to the image cache.
Related Methods

The pfQueue class provides a variety of other methods, described in Table 15-1, that
return information about the threads created to process the elements in the output buffer
of the pfQueue object.
Table 15-1

Thread Information

Method

Description

getServiceProcPID()

Returns the ID of the created thread.

pfGetGlobalQueueServiceProcPID()

Returns the ID of the nth thread.

getNumServiceProcs()

Returns the number of currently active threads.

pfGetNumGlobalQueueServiceProcs() Returns the number of processes that have been sproc’d
by all pfQueues.
pfGetGlobalQueueServiceProcQueue() Returns the pfQueue associated with a particular thread.
exitServiceProc()

Terminates a specific thread.

exitAllServiceProcs()

Terminates all pfQueue object threads.

pfQueue Modes
The pfQueue objects can run in one of two modes:
•

Nonsorting

•

Sorting

Either the elements in the queue are sorted according to some criteria specified by a
developer-supplied sorting function or not.
The sorting function is NULL and the sorting mode is nonsorting by default.

490

007-1680-080

Handling Queues

NonSorting Mode

In nonsorting mode, the sorted list and the output buffer are empty; all pfQueue
elements are in the input buffer. Processes append new input objects to the front of the
queue while (potentially) other processes read and remove pfQueue objects from the
other end of the queue.
A process can potentially read and remove all of the elements in a nonsorted queue.
Access to the elements is not random, however; it is sequential and ordered according to
FIFO.
Sorting the pfQueue

Multiple processes can add to the input buffer asynchronously. The objects remain
unsorted and separate from the sorted list and output buffer until the sorting function is
triggered. At that time, the following events occur:
1.

The objects in the input buffer are flushed into the sorted list.

2. The objects in the sorted list and the output buffer are resorted together.
To sort the elements in a pfQueue, you do the following:
1.

Specify a developer-supplied sorting function using pfQueue::setSortFunc().

2. Enable sorting by passing a non-NULL argument to pfQueue::setSortMode().
3. Specify the maximum and minimum number of values for the input and output of
the sorting function using pfQueue::setInputRange() and
pfQueue::setOutputRange().
Tip: You must specify the sorting function before enabling sorting; otherwise, sorting
remains disabled and pfQueue returns a warning.
The sorting function runs in a separate thread parallel to the function specified in the
argument of pfQueue. You can even specify that the sorting function run on a CPU
different from the one processing the pfQueue object, as described in “Running the Sort
Process on a Different CPU” on page 493.

007-1680-080

491

15: Managing Nongraphic System Tasks

In sorting mode, only those elements in the output buffer are available to processes.
Access to the elements in the output buffer is not random, but sequential and in a FIFO
order.
Sorting Function

The sorting function sorts, according to its own criteria, the elements in the sorted list and
the output buffer. To sort the queue, you must do the following:
•

Implement your own function to sort the pfQueue object.

•

Identify the function in your application using pfQueue::setSortFunc().

•

Make the function return a value of type that matches that of
pfQueueSortFuncType.

•

Make the function handle an input data structure of type pfQueueSortFuncData(),
defined as follows:
typedef struct {
pfList *sortList; //list of elements to sort
volatile int *inSize; //number of elements on input queue
volatile int *outSize; //number of elements on output queue
int inHi; // maximum number of elements at the input
int inLo; // minimum number of elements at the input
int outHi; // maximum number of elements at the output
int outLo; // minimum number of elements at the output
} pfQueueSortFuncData;

The actual data in the pfQueue object is maintained in a pfList, to which the
pfQueueSortFuncData structure points.
Input and Output Ranges

The range values work as triggers to start the sorting function, which sleeps otherwise.
For example, when the number of unprocessed inputs is greater than inHi, pfQueue calls
the sorting function to sort the pfQueue object.
You can set the minimum and maximum number of input and output elements entered
before the sort is triggered using the following methods:

492

•

pfQueue::setInputRange()

•

pfQueue::setOutputRange()

007-1680-080

Handling Queues

Table 15-2 shows the default range values:
Table 15-2

Default Input and Output Ranges

Range

Minimum

Maximum

Input

Output

The range values have no effect in nonsorting mode.
Triggering the Sort

The sorting function sleeps until one of the following conditions occurs:
•

The number of elements in the input buffer exceeds the input maximum range
value.

•

The number of elements in the output buffer drops below the output minimum
range value.

•

pfQueue::notifySortProc() is called.

By increasing the maximum number of values allowed in the input buffer, or reducing
the minimum number of values allowed in the output buffer, the sorting function is
potentially called fewer times.
Table 15-2 shows that, using default range values, the queue is sorted when three or more
elements are added to the input buffer or when two or less values remain in the output
buffer.
The pfQueue::notifySortProc() method is provided for those times when the queue
should be sorted without regard to the number of elements in the input or output buffers.
For example, if an element in the queue changes, it might be necessary to re-sort the
queue. If, for example, the elements are sorted alphabetically, the sort function should be
explicitly called when one of the elements is renamed.

Running the Sort Process on a Different CPU
You can run the sorting process on a different CPU from the one processing the pfQueue
by doing one the following:

007-1680-080

493

15: Managing Nongraphic System Tasks

•

Use getSortProcPID() to get the process ID of the sorting function and assigning the
process to run on a specified CPU with OpenGL Performer or operating system
utilities.

•

Use the pfuProcessManager provided in libpfutil. See the
pfuInitDefaultProcessManager(3) man page for more information.

High-Resolution Clocks
OpenGL Performer provides access to a high-resolution clock that reports elapsed time
in seconds to support for timing operations. To start a clock, call pfInitClock() with the
initial time in seconds—usually 0.0—as the parameter. Subsequent calls to pfInitClock()
reset the time to whatever value you specify. To read the time, call pfGetTime(). This
function returns a double-precision floating point number representing the seconds
elapsed from initialization added to the latest reset value.
The resolution of the clock depends on your system type and configuration. In most
cases, the resolved time interval is under a microsecond, and so is much less than the
time required to process the pfGetTime() call itself. Silicon Graphics Onyx, Crimson,
Indigo2, Indigo, and Indy™ systems all provide submicrosecond resolution. Newer
systems, including Silicon Graphics Onyx2, Silicon Graphics Onyx3, Silicon Graphics
Octane, Silicon Graphics Octane2, and Silicon Graphics O2 have even higher resolution
clocks and use the CYCLE_COUNTER functionality through the syssgi(2). On a
machine that uses a fast hardware counter, the first invocation of pfInitClock() forks off
a process that periodically wakes up and checks the counter for wrapping. This
additional process can be suppressed by using pfClockMode().
If OpenGL Performer cannot find a fast hardware counter to use, it defaults to the
time-of-day clock, which typically has a resolution between one and ten milliseconds.
This clock resolution can be improved by using fast timers. See the ftimer(1) man page
for more information on fast timers.
By default, processes forked after the first call to pfInitClock() share the same clock and
will all see the results of any subsequent calls to pfInitClock(). All such processes receive
the same time.
Unrelated processes can share the same clock by calling pfClockName() with a clock
name before calling pfInitClock(). This provides a way to name and reference a clock. By
default, unrelated processes do not share clocks.

494

007-1680-080

Memory Allocation

Video Refresh Counter (VClock)
The video refresh counter (VClock) is a counter that increments once for every vertical
retrace interval. There is one VClock per system. In systems where multiple graphics
pipelines are present, but not genlocked (synchronized, see the setmon(3) man page),
screen 0 is used as the source for the counter. A process can be blocked until a certain
count, or the count modulo some value (usually the desired number of video fields per
frame) is reached.
Table 15-3 lists and describes the pfVClock routines.
Table 15-3

pfVClock Routines

Routine

Action

pfInitVClock()

Initialize the clock to a value.

pfGetVClock()

Get the current count.

pfVClockSync()

Block the calling process until a count is reached.

When using pfVClockSync(), the calling routine is blocked until the current count
modulo rate is offset. The VClock functions can be used to synchronize several channels
or pipelines.

Memory Allocation
You can use OpenGL Performer memory-allocation functions to allocate memory from
the heap, from shared memory, and from data pools.
Note: On Microsoft Windows systems, all memory allocation is from the heap.

007-1680-080

495

15: Managing Nongraphic System Tasks

Table 15-4 lists and describes the OpenGL Performer shared-memory routines.
Table 15-4

Memory Allocation Routines

Routine

Action

pfInitArenas()

Create arenas for shared memory and semaphores.

pfSharedArenaSize()

Specify the size of a shared-memory arena.

pfGetSharedArena()

Get the shared-memory arena pointer.

pfGetSemaArena()

Get the shared-semaphore/lock arena pointer.

pfMalloc()

Allocate from an arena or the heap.

pfFree()

Release memory allocated with pfMalloc().

Allocating Memory With pfMalloc()
The pfMalloc() function can allocate memory either from the heap or from a shared
memory arena. Multiple processes can access memory allocated from a shared memory
arena, whereas memory allocated from the heap is visible only to the allocating process.
Pass a shared-memory arena pointer to pfMalloc() to allocate memory from the given
arena. pfGetSharedArena() returns the pointer for the arena allocated by pfInitArenas()
or NULL if the given memory was allocated from the heap. Alternately, an application
can create its own shared-memory arena; see the acreate(3P) man page for information
on how to create an arena.
To allocate memory from the heap, pass NULL to pfMalloc() instead of an arena pointer.
Under normal conditions pfMalloc() never returns NULL. If the allocation fails,
pfMalloc() generates a pfNotify() of level PFNFY_FATAL; so, unless the application has
set a pfNotifyHandler(), the application will exit.
Memory allocated with pfMalloc() must be freed with pfFree(), not with the standard C
library’s free() function. Using free() with data allocated by pfMalloc() will have
devastating results.
Memory allocated with pfMalloc() has a reference count (see “pfDelete() and Reference
Counting” in Chapter 1 for information on reference counting). For example, if you use
pfMalloc() to create attribute and index arrays, which you then attach to pfGeoSets using

496

007-1680-080

Memory Allocation

pfGSetAttr(), OpenGL Performer automatically tracks the reference counts for the
arrays; this allows you to delete the arrays much more easily than if you create them
without pfMalloc(). All the reference-counting routines (including pfDelete()) work
with data allocated using pfMalloc(). Note, however, that pfFree() does not check the
reference count before freeing memory; use pfFree() only when you are sure the data you
are freeing is not referenced.
The pfGetSize() function returns the size in bytes of any data allocated by pfMalloc().
Since the size of such data is known, pfCopy() also works on allocated data.
Although dtat allocated by pfMalloc() behaves in many ways like a pfObject (see
“Nodes” in Chapter 3), such data does not contain a user-data pointer. This omission
avoids requiring an extra word to be allocated with every piece of pfMalloc() data.
Note: All libpr objects are allocated using pfMalloc(); so, you can use pfGetArena(),
pfGetSize(), and pfFree() on all such objects. However, use pfDelete() instead of
pfFree() for libpr objects in order to maintain reference-count checking.

Shared Arenas
Note: Shared arenas are not pertinent to Microsoft Windows systems.
The pfInitArenas() function creates two arenas, one for the allocation of shared memory
with pfMalloc() and one for the allocation of semaphores and locks with usnewlock()
and usnewsema(). The arenas are visible to all processes forked after pfInitArenas() is
called.
Applications using libpf do not need to explicitly call pfInitArenas(), since it is
invoked by pfInit().
The shared memory arena can be allocated by memory-mapping either swap space
(/dev/zero, the default) or an actual disk file (in the directory specified by the
environment variable PFTMPDIR). The latter requires sufficient disk space for as much
of the shared memory arena as will be used, and disk files are somewhat slower than
swap space in allocating memory.

007-1680-080

497

15: Managing Nongraphic System Tasks

By default, OpenGL Performer creates a large shared memory arena (256 MB on IRIX and
64 MB on Linux). Though this approach makes large numbers appear when you run
ps(1), it does not consume any substantial resources, since swap or file system space is
not actually allocated until accessed (that is, until pfMalloc() is called).
Note: The following description applies only to IRIX systems.
Because OpenGL Performer cannot increase the size of the arena after initialization, an
application requiring a larger shared memory arena should call pfSharedArenaSize() to
specify the maximum amount of memory to be used. Arena sizes as large as 1.7 GB are
usually acceptable; but you may need to set the virtual-memory-use and memory-use
limits, using the shell limit command or the setrlimit() function, to allow your
application to use that much memory. To use arenas larger than 4 GB, you must use 64-bit
operation.
If you are having difficulties in creating a large arena, it could be due to fragmentation of
the address space from too many DSOs. You can reduce the number of DSOs you are
using by compiling some of them statically. You can also change the default address of
the DSOs by running the rqs(1) with a custom so_locations file.

Allocating Locks and Semaphores
An application requiring lockable pieces of memory should consider using pfDataPools,
described in the following section. Alternatively, when a lock or semaphore is required
in an application that has called pfInitArenas(), you can call pfGetSemaArena() to get
an arena pointer, and you can allocate locks or semaphores using usnewlock() and
usnewsema().

Datapools
Datapools, or pfDataPools, are also a form of shared memory, but they work differently
from pfMalloc(). Datapools allow unrelated processes to share memory and lock out
access to eliminate data contention. They also provide a way for one process to access
memory allocated by another process.
Any process can create a datapool by calling pfCreateDPool() with a name and byte size
for the pool. If an unrelated process needs access to the datapool, it must first put the

498

007-1680-080

Memory Allocation

datapool in its address space by calling pfAttachDPool() with the name of the datapool.
The datapool must reside at the same virtual address in all processes. If the default choice
of an address causes a conflict in an attaching process, pfAttachDPool() will fail. To
avoid this, call pfDPoolAttachAddr() before pfCreateDPool() to specify a different
address for the datapool.
Any attached process can allocate memory from the datapool by calling pfDPoolAlloc().
Each block of memory allocated from a datapool is assigned an ID so that other processes
can retrieve the address using pfDPoolFind().
Once you have allocated memory from a datapool, you can lock the memory chunk (not
the entire pfDataPool) by calling pfDPoolLock() before accessing the memory. This
locking mechanism works only if all processes wishing to access the datapool memory
use pfDPoolLock() and pfDPoolUnlock(). After a piece of memory has been locked
using pfDPoolLock(), any subsequent pfDPoolLock() call on the same piece of memory
will block until the next time a pfDPoolUnlock() function is called for that memory.
The pfDataPools are pfObjects; so, call pfDelete() to delete them. Calling
pfReleaseDPool() unlinks the file used for the datapool—it does not immediately free
the memory that was used or prevent further allocations from the datapool; it just
prevents processes from attaching to it. The memory is freed when the last process
referring to the datapool pfDelete() to remove it.

CycleBuffers
A multiprocessed environment often requires that data be duplicated so that each
process can work on its own copy of the data without adversely colliding with other
processes. pfCycleBuffer is a memory structure which supports this programming
paradigm. A pfCycleBuffer consists of one or more pfCycleMemories, which are
equally-sized memory blocks. The number of pfCycleMemories per pfCycleBuffer is
global, is set once with pfCBufferConfig(), and is typically equal to the number of
processes accessing the data.
Note: pfFlux replaces the functionality of pfCycleBuffer.
Each process has a global index, set with pfCurCBufferIndex(), which indexes a
pfCycleBuffer’s array of pfCycleMemories. When each process has a different index (and

007-1680-080

499

15: Managing Nongraphic System Tasks

its own address space), mutual exclusion is ensured if the process limits its
pfCycleMemory access to the currently indexed one.
The “cycle” term of pfCycleBuffer refers to its suitability for pipelined multiprocessing
environments where processes are arranged in stages like an assembly line and data
propagates down one stage of the pipeline each frame. In this situation, the array of
pfCycleMemories can be visualized as a circular list. Each stage in the pipeline accesses
a different pfCycleMemory and at frame boundaries the global index in each process is
advanced to the next pfCycleMemory in the chain. In this way, data changes made in the
head of the pipeline are propagated through the pipeline stages by “cycling” the
pfCycleMemories.

500

007-1680-080

Memory Allocation

2
Cull

pfC

1
Draw

ycl

uff

pfCycleMemories

0
Cull

0
App
1
Cull

2
Draw
1
App

Fra

0
Draw
2
App

n
Fra

n+1
Fra

Figure 15-2

n+2

pfCycleBuffer and pfCycleMemory Overview

Cycling the memory buffers works if each current pfCycleMemory is completely
updated each frame. If this is not the case, buffer cycling will eventually access a “stale”
pfCycleMemory whose contents were valid some number of frames ago but are invalid
now. pfCycleBuffers manage this by frame-stamping a pfCycleMemory whenever
pfCBufferChanged() is called. The global frame count is advanced with
pfCBufferFrame(), which also copies most recent pfCycleMemories into “stale”
pfCycleMemories, thereby automatically keeping all pfCycleBuffers current.

007-1680-080

501

15: Managing Nongraphic System Tasks

A pfCycleBuffer consisting of pfCycleMemories of nbytes size is allocated from
memory arena with pfNewCBuffer(nbytes, arena). To initialize all the pfCycleMemories
of a pfCycleBuffer to the same data call, pfInitCBuffer(). pfCycleMemory is derived
from pfMemory so you can use inherited routines like pfCopy() , pfGetSize(), and
pfGetArena() on pfCycleMemories.
While pfCycleBuffers may be used for application data, their primary use is as pfGeoSet
attribute arrays, for example, coordinates or colors. pfGeoSets accept pfCycleBuffers (or
pfCycleMemory) references as attribute references and automatically select the proper
pfCycleMemory when drawing or intersecting with the pfGeoSet.
Note: libpf applications do not need to call pfCBufferConfig() or pfCBufferFrame()
since the libpf routines pfConfig() and pfFrame() call these, respectively.

Asynchronous I/O (IRIX only)
A nonblocking file interface is provided to allow real-time programs access to disk files
without affecting program timing. The system calls pfOpenFile(), pfCloseFile(),
pfReadFile(), and pfWriteFile() work in an identical fashion to their IRIX counterparts
open(), close(), read(), and write().
When pfOpenFile() or pfCreateFile() is called, a new process is created using sproc(),
which manages access to the file. Subsequent calls to pfReadFile(), pfWriteFile(), and
pfSeekFile() place commands in a queue for the file manager to execute and return
immediately. To determine the status of a file operation, call pfGetFileStatus().

Error Handling and Notification
OpenGL Performer provides a general method for handling errors both within OpenGL
Performer and in the application. Applications can control error handling by installing
their own error-handling functions. You can also control the level of importance of an
error.

502

007-1680-080

Error Handling and Notification

Table 15-5 lists and describes the functions for setting notification levels.
Table 15-5

pfNotify Routines

Routine

Action

pfNotifyHandler()

Install user error-handling function.

pfNotifyLevel()

Set the error-notification level.

pfNotify()

Generate a notification.

The pfNotify() function allows an application to signal an error or print a message that
can be selectively suppressed. pfNotifyLevel() sets the notification level to one of the
values listed in Table 15-6.
Table 15-6

Error Notification Levels

Token

Meaning

PFNFY_ALWAYS

Always print regardless of notify level.

PFNFY_FATAL

Fatal error.

PFNFY_WARN

Serious warning.

PFNFY_NOTICE

Warning.

PFNFY_INFO

Information and floating point exceptions.

PFNFY_DEBUG

Debug information.

PFNFY_FP_DEBUG

Floating point debug information.

The environment variable PFNFYLEVEL can be set to override the value specified in
pfNotifyLevel(). Once the notification level is set via PFNFYLEVEL, it cannot be
changed by an application.
Once the notify level is set, only those messages with a priority greater than or equal to
the current level are printed or handed off to the user function. Fatal errors cause the
program to exit unless the application has installed a handler with pfNotifyHandler().
Setting the notification level to PFNFY_FP_DEBUG has the additional effect of trapping
floating point exceptions such as overflows or operations on invalid floating point
numbers. It may be a good idea to use a notification level of PFNFY_FP_DEBUG while

007-1680-080

503

15: Managing Nongraphic System Tasks

testing your application so that you will be informed of all floating-point exceptions that
occur.

File Search Paths
OpenGL Performer provides a mechanism to allow referencing a file via a set of path
names. Applications can create a search list of path names in three ways: the PFPATH
environment variable, the function pfFilePathv(), or the function pfFilePath(). (Note
that the PFPATH environment variable controls file search paths and has nothing to do
with the pfPath data structure.)
Table 15-7 describes the routines for working with pfFilePaths.
Table 15-7

pfFilePath Routines

Routine

Action

pfFilePath()
pfFilePathv()

Create a search path.

pfFindFile()

Search for the file using the search path.

pfGetFilePath()

Supply current search path.

You can specify a search path using pfFilePath(path), pfFilePathv(path0, path1, ..., pathn,
NULL), or with the environment variable PFPATH. You can specify any number of
directories using pfFilePath() and a maximum of 64 using pfFilePathv(). Colons
separate path names on IRIX and Linux and semicolons on Windows. Since
pfFilePathv() allows you to specify path names delimited by commas, it provides much
more economy in coding compared to the use of pfFilePath(), where you must employ
conditional code to accomodate cross-platform use.
Directories are searched in the order given, beginning with those specified in PFPATH,
followed by those specified by pfFilePath() or pfFilePathv(). Calling pfFilePath() or
pfFilePathv() a second time replaces the current path list rather than appending to it.
The function pfFindFile() searches the paths in PFPATH first, then those given in the
most recent pfFilePath() call; it returns the complete path name for the file if the file is
found. OpenGL Performer applications should use pfFindFile() (either directly or
through routines such as pfdLoadFile()) to look for input data files.

504

007-1680-080

File Search Paths

The pfGetFilePath() function returns the last search path specified by a pfFilePath() or
pfFilePathv() call. It does not return the path specified by the PFPATH environment
variable. If you want to find out that value, call getenv().

007-1680-080

505

Chapter 16

16. Dynamic Data

Making your data dynamic allows your scenes to change. Geometries can change
location, orientation, color, texture, or change into different things altogether. This
chapter explains how to create dynamic structures that can generate and manipulate
their own dynamic data using pfFlux, pfEngine, and pfFCS. A pfEngine computes the
changes and a pfFlux is a container for the output of the pfEngines.

pfFlux
A pfFlux is a container for dynamic data that enables multiprocessed generation and use
of data in the scene graph. A pfFlux internally consists of multiple buffers of data, each
of which is associated with a frame number. The numbering allows multiple processes to
each have a copy of the data containing the frame on which they are working. Multiple
reader processes can share a copy of the current results or use the frame of results that is
appropriate for that process. Figure 16-1 illustrates how pfFlux and processes use frame
numbers.

007-1680-080

507

16: Dynamic Data

Frame
12

pfFlux
Frame
8

Process
Frame 12
Figure 16-1

Frame
5

Frame
4

Process
Frame 7

pfFlux Buffer

Process
Frame 4

How pfFlux and Processes Use Frame Numbers

The pfFlux in the figure has four buffers. Each buffer shows the frame number associated
with it. The figure shows four processes reading from the pfFlux. The process with frame
12 reads from a buffer with frame 12—the newest available buffer and an exact match to
the process frame number. The process with frame number 7 reads from a buffer with
frame 5–the latest buffer that is not too new for frame 7. Both processes with frame 4
share a buffer with frame number 4—an exact match.
One popular use of a pfFlux is to manage the attributes of a pfGeoSet. Each of the
attribute lists of a pfGeoSet can independently be a pfFlux. This means that an
application can change the position of geometry vertices on the fly in a safe, multiprocess
manner.
Other uses include using a pfFlux as a matrix of a pfFCS node. This enables modifying
the transformation of some target geometry from an asynchronous process in a safe,
multiprocess manner.

Creating and Deleting a pfFlux
The pfNewFlux() funtion creates and returns a handle to a pfFlux. Each pfFlux buffer is
a pfFluxMemory. When you create a pfFlux, you specify the number of buffers it contains
as well as the size of each buffer. The pfFlux buffers are automatically allocated from the
same arena as the parent pfFlux. When creating a pfFlux, you must specify how many

508

007-1680-080

pfFlux

buffers this pfFlux contains. Instead of specifying an exact number, you may specify the
symbol PFFLUX_DEFAULT_NUM_BUFFERS. In this case, the system will pick an
appropriate number of buffers based on the multiprocess configuration of your
application (set by pfMultiprocess). You can globally redefine this default number of
buffers for successive creation of pfFluxes using pfFluxDefaultNumBuffers(). If you
change the default number of buffers, the effect takes place only for pfFluxes created after
the change.
Note: OpenGL Performer sets the default number of pfFlux buffers when pfConfig() is
called according to the number and type of processes requested with pfMultiprocess().
Generally, the number of buffers is equal to one more than the number of running
processes that typically might generate pfFlux data or use pfFluxed results. LPOINT
processes are not included in the count; but DBASE, ISECT, and COMPUTE processes
and additional graphic pipes, creating additional DRAW and possibly CULL processes,
do add to the default number of pfFlux buffers.
You can return the number of buffers in a pfFlux and the number of bytes in the buffers
using the pfGetFluxDefaultNumBuffers() and pfGetFluxDataSize() routines,
respectively.
To delete a pfFlux, as with all pfObjects, use pfDelete().

Initializing the Buffers
Generally, pfFlux buffers do not require initialization because they are fully recomputed
for every frame. One case where you would want to initialize them is when the data in
the buffers is static. For example, if you have a pfFlux of coordinates where most do not
change, rather than continuously updating this static data, it is far more efficient to
initialize the buffers. You cannot depend on the order in which you get pfFlux buffers at
run time and coordinates in the buffer cannot assume results from the previous frame. If
an element in the buffer is ever to be dynamic, it should always be recomputed.
OpenGL Performer provides two ways of initializing or setting the data held in the
pfFlux buffers:

007-1680-080

•

pfFluxInitData() to provide a template

•

pfFluxCallDataFunc() to provide a callback function

509

16: Dynamic Data

You can immediately initialize all pfFlux buffers by calling the pfFluxInitData() routine
and providing a template data buffer that will be directly copied into the pfFlux buffers.
In the argument of pfFluxCallDataFunc(), you pass a pointer to data and a pointer to a
callback function that operates on the buffers in a pfFlux. This function will be
immediately called on each buffer of the pfFlux.
An initialization callback can also be provided when the pfFlux is created as shown in
the following:
pfFlux *flux = new pfFlux(initFunc, PFFLUX_DEFAULT_NUM_BUFFERS);

The function will be called for each flux buffer. If the pfFlux is not created with enough
buffers and a new pfFluxMemory buffer must be created at run time, the callback
function will be called. For better and more reliable performance, you generally want to
ensure that you have given yourself enough buffers up front.

pfFlux Buffers
A pfFlux buffer is of type pfFluxMemory. Each buffer consists of the following:
•

Header

•

Data

The header portion consists of the following:
•

Pointer to the data portion of the pfFluxMemory

•

Frame number set automatically or explicitly by pfFluxFrame()

•

Set of flags, including read and write

The data portion of a pfFluxMemory contains one frame of information. The
pfGetFluxMemory() function returns a pointer to the data portion of a pfFluxMemory.
Figure 16-2 demonstrates these relationships.

510

007-1680-080

pfFlux

pfFluxMemory
Header

Data Buffer 0

pfFluxMemory
Header

Data Buffer 1

pfFluxMemory
Header

Data Buffer 2

pfFluxMemory
Header

Data Buffer n

pfFlux

Figure 16-2

pfFlux Buffer Structure

To return the parent pfFlux containing the buffer, use pfGetFFlux(), which returns NULL
if the specified data is not part of a pfFlux. This is useful to find out, for example, if an
attribute buffer of a pfGeoSet is a pfFlux.
At a given moment, a pfFlux buffer is either readable or writable, but never both. When
a buffer is writable, its data can be changed. When a buffer is readable, its data should
not be changed because there might be other processes reading that same buffer that
would then be immediately affected by any changes. For performance reasons, a buffer
marked readable is not locked.
Reading pfFlux Buffers

To get the current results for reading from a pfFlux, you can use the pfGetFluxCurData()
method with code similar to the following C example:
pfVec3 *cur_verts;
cur_verts = (pfVec3*)pfGetFluxCurData(flux);

007-1680-080

511

16: Dynamic Data

In this example, pfGetFluxCurData() does the following:
•

Finds a readable buffer in the pfFlux, flux, whose frame number is the closest to the
current flux frame number but not greater then the current flux frame number.

•

Returns a pointer, cur_verts, to that buffer.

If you have an old copy of a data buffer from a previous call to pfGetFluxCurData() and
now just want to update to a new version and no longer have a pointer to the parent flux,
you can use pfGetFluxWritableData() and provide your data pointer. For performance
reasons, it is better to save your pfFlux pointer and not depend on this convenience.
When pfGetFluxCurData() is called, it is expected to hold previous results of
computation. This computation might be done explicitly by your code in another process
or might be done automatically if the pfFlux is the destination of a pfEngine.
Additionally, the pfEngine computation might be triggered immediately if the pfFlux is
a pfEngine destination and if the pfFlux mode PFFLUX_ON_DEMAND mode is set by
the pfFluxMode() function and if the client pfEngine sources are dirty. pfEngine details
are discussed more later in this chapter.
Writing to pfFlux Buffers

When you want to write data to a pfFlux buffer, use the pfGetFluxWritableData()
function with code similar to the following C++ example:
pfVec3 *verts;
int i, num_verts;
verts = (pfVec3*)flux->getWritableData();
/* Set all verts to 1.0, 2.0, 3.0 */
num_verts = flux->getDataSize() / sizeof(pfVec3);
for (i = 0; i < num_verts; i++)
verts[i].set(1.0f, 2.0f, 3.0f);
flux->writeComplete();

When pfGetFluxWritableData() is called, pfFlux searches for the buffer whose frame
number is equal to the pfFlux frame number. There are three possible results:
•

512

If there is a match and the buffer is writable, pfGetFluxWritableData() returns a
pointer to that buffer.

007-1680-080

pfFlux

•

If there is a match but the buffer is readable, pfGetFluxWritableData() causes one of
the following actions to occur, according to whether the PFLUX_WRITE_ONCE
mode specified in pfFluxMode() is set to the following:
PF_ON

The pfGetFluxWritableData() function returns NULL if there is
already a readable buffer with a frame number that matches the
current flux frame number.

PF_OFF

The pfGetFluxWritableData() function returns a pointer to the
readable data buffer.

If there is no match, an unused buffer is made writable and its frame buffer number
is set to that of the current pfFlux frame number.

If pfGetFluxWritableData() is called again for the same frame and the
PFFLUX_WRITE_ONCE mode is PF_ON (off by default), information is not copied
again into the same buffer; instead, NULL is returned. In this way, the mode can prevent
modified data from being destroyed by a second call to pfGetFluxWritableData() and
the NULL return value can be used to avoid needlessly recomputing unused data.
The pfFluxWriteComplete() function should be called when computation for a writable
pfFlux buffer is complete. This method changes the specified buffer from writable to
readable.
Note: OpenGL Performer computes a cache of geometric attributes for each pfGeoSet. It
uses this cache when intersecting line segments with the pfGeoSet. When using a pfFlux
as the PFGS_COORD3 attribute of a pfGeoSet, the cache is not notified if pfFlux changes;
so, intersection is incorrect. In order to avoid this problem, the application should always
disable caching in pfGeoSets with pfFluxed PFGS_COORD3 attributes. Use the pfGeoSet
function pfGSetIsectMask(gset, PFTRAV_IS_UNCACHE, ...) to disable caching.

Coordinating pfFlux and Connected pfEngines
The pfFluxes maintain pointers to the following:
•

pfEngines, called source engines, whose destinations are this pfFlux.

•

A list of pfEngines, called client engines, that use this pfFlux as a source.

To return a handle to these engines, use pfGetFluxSrcEngine() and
pfGetFluxClientEngine(), respectively.

007-1680-080

513

16: Dynamic Data

To return the number of connected source and client pfEngines, use
pfGetFluxNumSrcEngines() and pfGetFluxNumClientEngines(), respectively.
The pfFluxWriteComplete() function triggers client pfEngine evaluation according to
whether the PFFLUX_PUSH mode specified in pfFluxMode() is set to the following:
PF_ON

The pfEngineEvaluate() function is performed on its client pfEngines to
push the results through a chain of computation.

PF_OFF

The pfEngineSrcChanged() function is performed on its clients to tell
those pfEngines that they have dirty sources.

Triggering pfFlux Evaluation

pfFlux evaluation is really a convenient one-step combined evaluation of a source
pfEngine combined with getting and completing a writable buffer of data. Triggering the
evaluation of pfFlux may trigger the evaluation of the pfFlux’s source pfEngines,
particularly if the PFFLUX_ON_DEMAND mode is set to PF_ON.
To explicitly trigger a pfFlux, the evaluate method can be used with a mask that is
provided and passed through evaluation chains to potentially limit which pfFluxes in a
chain are evaluated. For a pfFlux to be evaluated, both of the following conditions must
be satisfied:
•

The bits in the current mask must match any of the bits in the evaluation mask
associated with a pfFlux.

•

The pfFluxEvaluate() or pfFluxEvaluateEye() function is called.

The mask is a bitmask that you use to trigger the evaluation of the source pfEngines of a
pfFlux. The functions pfFluxMask() and pfGetFluxMask() set and get, respectively, the
evaluation mask of a pfFlux. The default mask is PFFLUX_BASIC_MASK. Masks enable
selective evaluation of pfFlux source pfEngines.
The pfFluxEvaluate() function triggers an evaluation of pfFlux source pfEngines if any
of the bits in the current mask match any of the bits in the evaluation mask of the pfFlux.
The pfFluxEvaluateEye() function is the same as pfFluxEvaluate() but also makes it easy
to pass the current eyepoint through a chain of computation. This is a particularly
common case in scene graphs and one very common use is morphing level of detail
based on distance from the current eyepoint.

514

007-1680-080

pfFlux

Calling pfFluxEvaluate() or pfFluxEvaluateEye() is the equivalent of calling
pfEngineEvaluate() on the source engines followed by calling pfFluxWriteComplete()
on the pfFlux.
For more information about pfEngine, see “pfEngine” on page 522.

Synchronized Flux Evaluation
There are times when you want to ensure that the dynamic data for a given frame in a
number of pfFluxes becomes readable at precisely the same time. Particularly when
computations are so complicated that they must be completed over multiple frames, you
might want the computations of vertices for the different pfGeoSets for a given frame to
be made usable at the same time. Consider the following example:
Suppose that you are using pfASD and you need shapes comprised of multiple
pfGeoSets to move together in response to morphing terrain. One pfGeoSet might
encapsulate the texture of a roof, another a wall, another a window, and another a door.
Since a pfGeoSet can only encompass one texture, each of these parts of a house must
remain separate pfGeoSets whose data must be encapsulated in separate pfFluxes. If, for
rendering, the roof of one frame was used with the wall of another, it might appear that
the house is breaking apart.
OpenGL Performer uses sync flux groups to ensure that the same frame of data is being
used from all fluxes in a group.
Synchronizing pfFluxes with Flux Sync Groups

You can add one or more pfFluxes to a flux sync group using the pfFlux
pfFluxSyncGroup() function and providing the integer identifier of the desired group.
To get the flux sync group identifier of a pfFlux, use the pfGetFluxSyncGroup() function.
Note: OpenGL Performer does not maintain a list of pfFluxes in a flux group; instead, an
internal field in pfFlux identifies the flux sync group to which it belongs. For this reason,
there is no way to get a list from OpenGL Performer of all of the fluxes in a sync group.
For convenience, a flux sync group can be identified by a string name, set with
pfGetFluxSyncGroupName(). The unsigned integer identifier can be obtained from the
string name using pfGetFluxNamedSyncGroup(). The pfGetFluxNamedSyncGroup()

007-1680-080

515

16: Dynamic Data

function also automatically generates a new sync group identifier for a new name. The
identifier of the first flux sync group you create is automatically one.
Note: Once you name a sync group, the name cannot be changed.
The pfGetFluxNumNamedSyncGroups() function returns the number of named sync
groups.
Enabling a Flux Sync Group

To enable group synchronization of the pfFluxs in the flux sync group, enable the sync
group using pfFluxEnableSyncGroup(). You can disable group synchronization using
pfFluxDisableSyncGroup().
The pfGetFluxEnableSyncGroup() function returns whether or not a sync group is
enabled.
Initially, all pfFluxes are all part of flux sync group 0, which can never be enabled.
Evaluating a Synchronized Flux Group

After all of the pfFluxes in an enabled flux sync group are calculated, you must call
pfFluxSyncGroupReady() to specify that the fluxes are ready to be read.
When pfFrame() is called, pfFluxSyncComplete() is called on flux sync groups, which
does the following:
•

Marks the writable buffers with the current flux frame number.

•

Makes their writable buffers readable.

Figure 16-3 is a timing diagram that shows a typical use for sync groups.

516

007-1680-080

pfFlux

Calls to
pfFlux::writeComplete

Call to
pfFlux::syncGroupReady

Asynchronous process frame 7
Delay until next pfFrame
8

APP/CULL/DRAW frame boundaries
Data is available for
all fluxes.
Figure 16-3

Timing Diagram Showing the Use of Sync Groups

The diagram shows calls to pfFluxWriteComplete() on three pfFluxes during the course
of a single asynchronous frame 7 (for example, a COMPUTE frame). Although the main
processes APP, CULL, and DRAW run in frame numbers higher than 7, they cannot
access the pfFlux results of the asynchronous process until the first pfFrame() after the
call to pfFluxSyncGroupReady(). APP, CULL, and DRAW start seeing the results at
frame 13. In this way an application can synchronize multiple results generated by an
asynchronous frame. They all become visible at exactly the same frame boundary.
Note: Using sync groups adds some processing to pfFrame. When a pfFlux is not a
member of a sync group, OpenGL Performer spends no per-frame processing on it. The
only time a pfFlux consumes CPU time is during its API calls. However, when a pfFlux
is on a sync group, OpenGL Performer incurs extra processing time at frames where the
pfFlux data becomes available to the APP process (the pfFrame() following a call to
pfFluxSyncGroupReady()).

007-1680-080

517

16: Dynamic Data

Fluxed Geosets
Most often you use dynamic data to change the attributes, location, or the orientation of
geometries. Morphing geometries, for example, is a matter of repositioning the vertices
of a geometry.
You can, however, use dynamic data to change the higher level description of geometries.
You might, for example, create a geometry editor that adds or subtracts triangles to and
from geometries.
Note: While OpenGL Performer is set up to flux any object derived from pfMemory, only
pfGeode and pfGeoSet are currently modified to accept fluxed and unfluxed forms
without any special pfFlux method.
To dynamically change a pfGeoSet, you must operate on the data held in the pfFlux
buffers usually when the buffers are initialized.
Example 16-1 shows how to turn the data portion of a pfFluxMemory into a fluxed
pfGeoSet using pfFluxedGSetInit().
Example 16-1

Fluxed pfGeoSet

int make_fluxed_gset(pfFluxMemory *fmem)
{
pfGeoSet *gset;
pfVec3 *coords;
if (fmem == NULL)
return pfFluxedGSetInit(fmem);
pfFluxedGSetInit(fmem);
gset = (pfGeoSet*)fmem->getData();
gset->setPrimType(PFGS_TRIS);
... // finish initializing pfGeoSet
return 0;
}

518

007-1680-080

pfFlux

main()
{
pfFlux *fluxed_gset;
pfGeoSet *gset;
pfInit();
pfMultiprocess(PFMP_DEFAULT);
pfConfig();
fluxed_gset = new pfFlux(make_fluxed_gset,
PFFLUX_DEFAULT_NUM_BUFFERS);
gset = (pfGeoSet*)fluxed_gset->getCurData();
...
}

Fluxed Coordinate Systems
A pfFCS is similar to a pfDCS node in that both nodes contain dynamic data. In addition
to pfDCS functionality, however, a pfFCS uses a pfFlux to hold its matrix. This fluxed
matrix can then be computed by a pfEngine, potentially asynchronously. This is the
structure shown in Figure 16-4.

pfGroup

pfEngine

pfFlux

pfFCS

pfEngine
pfGeoSet
pfMemory

pfFlux

Figure 16-4

007-1680-080

pfEngine Driving a pfFlux That Animates a pfFCS Node

519

16: Dynamic Data

In this figure, the pfEngine performs calculations on the data input from multiple
sources, including a pfMemory, a container for static data, a pfFlux, and an additional
source pfEngine. The main pfEngine sends its resulting matrix to a pfFlux. The pfFlux is
attached to the pfFCS with the pfFCS pfFlux() function. This flux then can do the
following:
•

Trigger the pfEngine to directly recompute its data if in PFFLUX_ON_DEMAND
mode when the pfFCS calls pfGetFluxCurData() on the flux

•

Use frame accurate results with other fluxes if a member of a flux sync group

•

Contain the matrix for the pfFCS node to produce a transformed coordinate system
for the current frame for children of the node in the scene graph.

Since a pfFCS node is of type pfGroup, you can connect many nodes to it. This
functionality is valuable if many geometries need to share a transformation, such as the
moving limbs of a walking character whose overall location is changing every frame.
Unlike pfDCS nodes, pfFCS nodes do not detect changes in their matrix. A change to the
matrix can move the children of a pfFCS node and change their visibility. This means that
changing a pfFCS matrix may result in a visible node being culled out or invisible nodes
being drawn. Both potential results are undesirable. In order to avoid them, an
application has two options:
•

If the objects under the pfFCS node move in some known and confined space, the
application can pre-calculate the bounding sphere of that space and set the
bounding sphere of the pfFCS node statically to that sphere. In this way, changes to
the pfFCS matrix will not change the visibility of the objects under the pfFCS node.
The smaller the extent of the motion, the more efficient this method is. If there is no
prior knowledge of the extent, this method is very wasteful because the application
has to set a very large bounding sphere that is always visible.

•

In the APP process, change the bounding sphere of the pfFCS node every frame.
This is not very desirable. If an application changes the bounding sphere every
frame, it may just as well use a pfDCS node instead of a pfFCS node.

For an example of pfEngine, pfFlux, and pfFCS use, see fcs_engine.C in
sample/pguide/libpf/C++.

520

007-1680-080

pfFlux

Replacing pfCycleBuffer with pfFlux
Prior to version 2.2 of OpenGL Performer, the customary way of manipulating dynamic
data was to use a pfCycleBuffer. Morphing was accomplished using pfCycleBuffer and
pfMorph. While pfCycleBuffer and pfMorph are still supported for compatibility, they
are obsoleted by pfEngine and pfFlux.
Note: If your applications do not contain pfCycleBuffer or pfMorph, skip to the next
section. This section explains how to replace pfCycleBuffer and pfMorph with pfFlux
and pfEngine.

pfFlux Differences

A pfFlux is similar to, but far more powerful than, a pfCycleBuffer:
•

Where pfCycleBuffer is unaware of the nodes driving it, pfFlux is aware of its
parent nodes. When attached to a pfEngine, for example, a pfFlux can trigger a
pfEngine to recompute its output.

•

A pfFlux provides a mechanism for updating data in processes that are completely
asynchronous to the main APP, CULL, and DRAW pipeline stages.

Converting to pfFlux

To convert from pfCycleBuffer to pfFlux, use code similar to the following:
/* Replace pfCyclebuffer creation with pfFlux: */
pfFlux *flux = pfFlux(size, PFFLUX_DEFAULT_NUM_BUFFERS);
/* replace getting of read-only data
* pfCBufGetCurData() for read becomes:
*/
curData = flux->getCurData();
/* replace getting of data to edit.*/
/* get writable buffer BEFORE editing data */
data = flux->getWritableData();
/* ... edit data */

007-1680-080

521

16: Dynamic Data

/* declare data edited after editing for the frame is done.
* Replace pfCBufferChanged(pfCycleBuffer *cbuf) becomes:
*/
flux->WriteComplete();

pfEngine
A pfEngine performs calculations. The source can either be static data, such as a
pfMemory, or dynamic data, such as a pfFlux. The destination of a pfEngine is fed into a
pfFlux, as shown in Figure 16-4 on page 519.

Creating and Deleting Engines
The constructor for pfEngine requires that you specify the computation type of pfEngine
you are creating. OpenGL Performer provides many types of engines, each performing
a different calculation on the input data. Table 16-1 describes the engine types:
Table 16-1

522

pfEngine Types

Engine

Description

PFENG_SUM

Sums the input array of floats into a destination array of floats.

PFENG_MORPH

Morphs between the input data

PFENG_BLEND

Is a lightweight version of PFENG_MORPH.

PFENG_TRANSFORM

Translates a set of data with a matrix operation.

PFENG_ALIGN

Generates a translation matrix from the input data.

PFENG_MATRIX

Generates a matrix of type rotation, translation, scale,
non-uniform scaling, or combine, using pre or post multiplication.

PFENG_ANIMATE

Same as PFENG_MATRIX, except that the sources are arrays of
animation frames.

PFENG_BBOX

Computes the bounding box of the input array of data.

PFENG_TIME

Takes a time, in seconds, and makes it into a frame number that is
useful in driving the frame source of PFENG_MORPH,
PFENG_BLEND, and PFENG_ANIMATE.

007-1680-080

pfEngine

Table 16-1

pfEngine Types (continued)

Engine

Description

PFENG_STROBE

Switches iteration sets of floats, called items, between an on and off
based on time.

PFENG_USER_FUNCTION

User-defined function.

The pfGetEngineFunction() function returns the pfEngine type. These engine types are
described in further detail in “Setting Engine Types and Modes” on page 523.
To delete a pfEngine, as with all pfObjects, use pfDelete().

Setting Engine Types and Modes
Table 16-1 lists the different types of pfEngines supplied by OpenGL Performer. Many of
the pfEngines have different modes of operation. For example, PFENG_ANIMATE offers
the following modes:
•

PFENG_ANIMATE_ROT

•

PFENG_ANIMATE_TRANS

•

PFENG_ANIMATE_SCALE_UNIFORM

•

PFENG_ANIMATE_SCALE_XYZ

•

PFENG_ANIMATE_BASE_MATRIX

All of these modes specify what the engine changes. PFENG_ANIMATE_ROT, for
example, specifies that the engine rotates a geometry, PFENG_ANIMATE_TRANS
translates a geometry, and PFENG_ANIMATE_SCALE_UNIFORM scales a geometry.
The pfEngineMode() function sets the mode value; pfGetEngineMode() returns the
mode value.
The following sections describe the engine types and their mode values, if any.
PFENG_SUM Engines

A PFENG_SUM engine adds arrays of floats to form a destination array of floats. One use
for PFENG_SUM is aligning geometries to height, such as buildings, to a pfASD terrain.

007-1680-080

523

16: Dynamic Data

Since PFENG_SUM can have as few as one source, it can be used to simply copy data
from one location to another.
PFENG_MORPH Engines

Morphing is the smooth transition from one appearance to another. The effect is achieved
through the reorientation of the vertices in pfGeoSets. Morphing can refer to geometries
and their attributes. For example, you could “morph” the following:
•

Color to simulate a flickering fire.

•

Texture coordinates to simulate rippling ocean waves.

•

Coordinates to make a 3D model of a face smile or frown.

A PFENG_MORPH engine sets the destination of the pfEngine to a weighted sum of its
sources. To specify the weighting, you use one of the following:
•

PFENG_MORPH_WEIGHTS

•

PFENG_MORPH_FRAME

The token PFENG_MORPH_WEIGHTS contains an array of floats or weighting values.
PFENG_MORPH_SRC(n) is also an array; there is one element for each pfEngine source.
To create the morph weighting, element 0 of PFENG_MORPH_WEIGHTS is multiplied
by PFENG_MORPH_SRC(0); element 1 is multiplied by PFENG_MORPH_SRC(1), and
so on.
PFENG_MORPH_FRAME contains a single float. This float specifies the weighting
between two pfEngine sources, PFENG_MORPH_SRC(n). The integer portion of the
float specifies the first of the two consecutive sources, and the decimal portion of the float
specifies the weighting between those sources. For example, a
PFENG_MORPH_FRAME of 3.8 would mean PFENG_MORPH_SRC(3) * 0.2 +
PFENG_MORPH_SRC(4) * 0.8.
For an example of a morph engine, see morph_engine.C in
sample/pguide/libpf/C++.
PFENG_BLEND Engines

A PFENG_BLEND engine is a lightweight version of PFENG_MORPH. PFENG_BLEND
sets the pfEngine destination to a weighted sum of elements of the pfEngine sources.

524

007-1680-080

pfEngine

To specify the weighting, you use one of the following:
•

PFENG_BLEND_WEIGHTS

•

PFENG_BLEND_FRAME

These weighting mechanisms work identically to those in “PFENG_MORPH Engines,”
with the following exception: the source, PFENG_BLEND_SRC(n), should contain
iteration elements. Each element is a set of floats, called items. The number of elements in
each set is called a stride. So, the iteration items for pfEngine source 0 begin at items
number 0 * stride; for source 1, at items number 1 * stride, and so on.
PFENG_BLEND_WEIGHTS contains an array of floats, one for each of the pfEngine
sources, PFENG_BLEND_SRC(n).
PFENG_BLEND_WEIGHTS(0) is multiplied by the items starting at
PFENG_BLEND_SRC(0*stride); PFENG_BLEND_WEIGHTS(1) is multiplied by the
items starting at PFENG_BLEND_SRC(1*stride) and so on.
PFENG_BLEND_FRAME contains a single float. This float specifies the weighting
between two consecutive PFENG_BLEND_SRC(n) sources. The integer portion of the
float specifies the first of the two consecutive sources, and the decimal portion of the float
specifies the weighting between those sources. For example, a PFENG_BLEND_FRAME
of 3.8 would mean PFENG_BLEND_SRC(3) * 0.2 + PFENG_BLEND_SRC(4) * 0.8.
For an example of a blend engine, see blend_engine.C in
sample/pguide/libpf/C++.
PFENG_TRANSFORM Engines

A PFENG_TRANSFORM engine transforms the PFENG_TRANSFORM_SRC(n) array of
floats by the matrix contained in PFENG_TRANSFORM_MATRIX.
PFENG_ALIGN Engines

The following describes the PFENG_ALIGN engines:
•

007-1680-080

PFENG_ALIGN generates an alignment matrix based on the sources. One use for
PFENG_ALIGN is to align moving objects, such as vehicles, to a pfASD.

525

16: Dynamic Data

•

PFENG_ALIGN_POSITION determines the translational portion of the matrix. If
PFENG_ALIGN_POSITION is NULL, the translational portion of the matrix is set
to all zeros.

•

PFENG_ALIGN_NORMAL and PFENG_ALIGN_AZIMUTH determine the
rotation portion of the matrix. If either are NULL, the rotation portion of the matrix
is set to all zeros.

PFENG_MATRIX Engines

A PFENG_MATRIX engine generates a matrix based on its sources. Applied to vertex
coordinates of pfGeoSets, these routines provide the same mathematical effect as using
pfDCSs in a scene graph. However, these pfEngines use the host to compute the vertices
and, thus, eliminate the need for matrices to be evaluated in the graphics pipeline. This
trade-off might produce a faster rendering frame rate if there are sufficient host CPU
resources for the computation involved.
The following tokens specify the kind of action performed by the PFENG_MATRIX
engine:
PFENG_MATRIX_RO
Rotates a geometry according to heading, pitch, and roll values; the
equivalent is pfDCSRot().
PFENG_MATRIX_TRANS
Transforms a geometry; the equivalent is pfDCSTrans().
PFENG_MATRIX_SCALE_UNIFORM
Uniformly scales a geometry; the equivalent is pfDCSScale().
PFENG_MATRIX_SCALE_XYZ
Non-uniformly scales a geometry; the equivalent is
pfDCSScaleXYZ(x,y,z).
PFENG_MATRIX_BASE_MATRIX
Contains a pfMatrix that is either pre- or post-multiplied against the
matrix generated by the other sources, depending on the
PFENG_MATRIX_MODE mode set by the pfEngineMode() function;
PF_OFF specifies pre-multiplication; PF_ON specifies
post-multiplication.
Any or all of the sources can be NULL, in which case they have no effect on the resulting
matrix.

526

007-1680-080

pfEngine

PFENG_ANIMATE Engines

A PFENG_ANIMATE engine animates a matrix based on its sources. This engine type is
similar to PFENG_MATRIX, but instead of using single values for rotation, translation,
and scaling, PFENG_ANIMATE uses arrays of values.
To specify the weighting, you use one of the following:
•

PFENG_ANIMATE_WEIGHTS

•

PFENG_ANIMATE_FRAME

PFENG_ANIMATE_WEIGHTS contains a float for each of the values in the rotation,
translation, and scale sources. PFENG_ANIMATE_SRC(n) is also an array; there is one
element for each pfEngine source.
To create an animation, element zero of PFENG_ANIMATE_WEIGHTS is multiplied by
PFENG_ANIMATE_SRC(0); element one is multiplied by PFENG_ANIMATE_SRC(1),
and so on.
PFENG_ANIMATE_FRAME contains a single float. This float specifies the weighting
between two of the values in the rotation, translation, and scale sources. The integer
portion of the float specifies the first of the two consecutive values, and the fractional
portion of the float specifies the weighting between those values. For example, a
PFENG_ANIMATE_FRAME of 3.8 would mean PFENG_ANIMATE_SRC(3) * 0.2 +
PFENG_ANIMATE_SRC(4) * 0.8.
The following tokens specify the kind of action performed by the PFENG_ANIMATE
engine:
PFENG_ANIMATE_ROT
Rotates a geometry according to heading, pitch, and roll values; the
equivalent is pfDCSRot().
PFENG_ANIMATE_TRANS
Transforms a geometry; the equivalent is pfDCSTrans().
PFENG_ANIMATE_SCALE_UNIFORM
Uniformly scales a geometry; the equivalent is pfDCSScale().
PFENG_ANIMATE_SCALE_XYZ
Non-uniformly scales a geometry; the equivalent is pfDCSScaleXYZ(x,
y, z).

007-1680-080

527

16: Dynamic Data

PFENG_ANIMATE_BASE_MATRIX
Contains a pfMatrix that is either pre- or post-multiplied against the
matrix generated by the other sources depending on the
PFENG_MATRIX_MODE mode set by the pfEngine function
pfEngineMode().
Any or all of the sources can be NULL, in which case they have no effect on the resulting
matrix.
For an example of an animate engine, see fcs_animate.C in
sample/pguide/libpf/C++.
For more information about animation, see “Animating a Geometry” on page 532.
PFENG_BBOX Engines

A PFENG_BBOX engine generates a bounding box that contains the coordinates of the
pfEngine source, PFENG_BBOX_SRC.
PFENG_TIME Engines

A PFENG_TIME engine takes a time in seconds and converts it to a frame number, which
can drive the sources of the following engine types: PFENG_MORPH, PFENG_BLEND,
and PFENG_ANIMATE.
PFENG_TIME_TIME is the source time in seconds. This is usually connected to the
pfFlux returned from pfGetFluxFrame().
PFENG_TIME_SCALE contains four floats that are used to modify the incoming time:
•

PFENG_TIME_SCALE[0] is an initial offset.

•

PFENG_TIME_SCALE[1] is a scale factor.

•

PFENG_TIME_SCALE[2] is a range.

•

PFENG_TIME_SCALE[3] is a final offset.

The PFENG_TIME_MODE mode, set with pfEngineMode(), determines how the
destination number moves between its start and end point. The two mode values include
the following:

528

007-1680-080

pfEngine

•

PFENG_TIME_CYCLE makes the destination number go from start to end then
restarts again at the beginning.

•

PFENG_TIME_SWING makes the destination number go back and forth between
the start and end.

Note: This mode is related to the interval mode of a pfSequence.
If the PFENG_TIME_TRUNC mode is set to PF_ON, the result is truncated.
PFENG_STROBE Engines

A PFENG_STROBE engine switches iteration sets of floats, called items, between ON and
OFF based on time. One use of PFENG_STROBE is light point animations.
PFENG_STROBE_TIME is the source time in seconds. This data is usually connected to
the pfFlux returned from pfGetFluxFrame().
PFENG_STROBE_TIMING contains iterations sets of three floats:
•

PFENG_STROBE_TIMING[n*stride + 0] is the ON duration.

•

PFENG_STROBE_TIMING[n*stride + 1] is the OFF duration.

•

PFENG_STROBE_TIMING[n*stride + 2] is an offset.

PFENG_STROBE_ON contains iteration sets of floats, called items.
PFENG_STROBE_OFF contains iteration sets of floats, called items. If
PFENG_STROBE_OFF is NULL, all off states are 0.0.
For an example of animation using a strobe engine, see strobe_engine.C in
sample/pguide/libpf/C++.
PFENG_USER_FUNCTION Engines

When a pfEngine is of type PFENG_USER_FUNCTION, you specify the function of the
pfEngine using pfEngineUserFunction(). The pfGetEngineUserFunction() function
returns the pfEngine type.

007-1680-080

529

16: Dynamic Data

For an example of animation using a user-defined engine, see user_engine.C in
sample/pguide/libpf/C++.

Setting Engine Sources and Destinations
The pfEngine sources can be any number of objects, including pfFluxes, pfMemorys, and
pfEngines. The sources provide the input data for the pfEngine. pfEngine destinations
are pfFluxes, which contain the pfEngine output.
The pfEngineSrc() and pfEngineDst() functions set the pfEngine sources and
destination, respectively; pfGetEngineSrc() and pfGetEngineDst() return the pfEngine
sources and destination, respectively. Function pfGetEngineNumSrcs() returns the
number of sources. A pfEngine can only have one destination.

Setting Engine Masks
The pfEngine masks are bit masks that provide a means of selectively triggering
pfEngines. Only those pfEngines that have masks that match the evaluation mask can be
triggered, as shown in the evaluation function:
pfEngineEvaluate(int mask)

The pfEngineMask() and pfGetEngineMask() functions set and return pfEngine masks,
respectively.

Setting Engine Iterations
You can make pfEngines repeat their calculations, called iterations, when operating on an
array of data instead of on a single piece of data. You also specify the unit of data, called
an item, for example, a vector would have three items per unit. For example, if you want
to add two arrays of 100 pfVec3s each, you set iteration to 100 and item to 3.
The pfEngineIterations() and pfGetEngineIterations() functions set and get iterations,
respectively.

530

007-1680-080

pfEngine

Setting Engine Ranges
There are times when you might like to reduce computation needs by not evaluating
engines, for example, when the eyepoint is far from the geometry being animated by a
pfEngine.
You establish the location of an animated geometry and an area located around the
geometry using pfEngineEvaluationRange(). Only when the eye position is within that
area can the pfEngine be evaluated.
The range functionality is only enabled when the PFENG_RANGE mode, created by
pfEngineMode(), is set to PF_ON; the default is PF_OFF.

Evaluating pfEngines
To evaluate a pfEngine, you use one of the forms of pfEngineEvaluate() or
pfFluxEvaluate().
For more information about pfFluxEvaluate(), see “Triggering pfFlux Evaluation” on
page 514.
The two forms of the evaluate functions are the following:
pfEngineEvaluate(pfEngine* _engine, int mask)
pfEngineEvaluateEye(pfEngine* _engine, int mask, pfVec3 eye_pos)
pfFluxEvaluate(pfFlux* _flux, int mask)
pfFluxEvaluateEye(pfFlux* _flux, int mask, pfVec3 eye_pos)

For more information about mask, see “Setting Engine Masks” on page 530.
The second form of the function specifies the location of the viewer. The engine is only
evaluated if the location of the viewer, eye_pos, is within the range of the pfEngine, set by
pfEngineEvaluationRange(). For more information about the range of a pfEngine, see
“Setting Engine Ranges” on page 531.
Note: The eye position has no effect on evaluation of the pfEngine if the
PFENG_RANGE_CHECK mode is PF_OFF, the default.

007-1680-080

531

16: Dynamic Data

Animating a Geometry
To animate a geometry using a combination of pfFlux, pfFCS, and pfEngines nodes, use
the following guidelines:
1.

Initialize and populate the pfFluxes.

2. Connect pfMemory or pfFluxes as the sources of a pfEngine.
3. Connect a pfFlux to the output destination of the pfEngine.
4. Connect the pfFlux containing the output of the pfEngine to the scene graph in one
of three ways:
•

Connect it as an attribute of a pfGeoSet using pfGSetAttr().

•

Connect it as the bounding box of a pfGeoSet using pfGSetBound(), where the
mode argument is set to PFBOUND_FLUX.

•

Connect it directly to a pfFCS using pfFCSFlux().

5. Set up any needed flux sync groups for synchronizing transformations.
This scenario is the simplest set up; it is represented graphically in Figure 16-4 on
page 519. More complicated scenarios include chaining pfEngines together or running
multiple geometries off of one pfFCS node.
The following C++ code sample provides an implementation of the animation
procedure:
Example 16-2

Connecting Engines and Fluxes

// create the nodes
pfFlux *myData1 = new pfFlux(100 * sizeof(pfVec3));
pfFlux *myData2 = new pfFlux(100 * sizeof(pfVec3));
pfEngine *myEngine = new pfEngine(PFENG_SUM);
pfFlux *engineOutput = new pfFlux(100 * sizeof(pfVec3));
pfFCS myFCS = new pfFCS();
pfGeode myGeode = new pfGeode();
// initialize and populate the flux nodes
myData1->init();
myData2->init();
// attach the pfFlux nodes as the source of the pfEngine
myEngine->setSrc(0, myData1, 0, 3);

532

007-1680-080

Animating a Geometry

myEngine->setSrc(0, myData12, 0, 3);
// attach a pfFlux to the output of the pfEngine
myEngine->setDst(engineOutput, 0, 3);
myEngine->iterations(100, 3);
// connect the pfFlux output node to the scenegraph
myFCS->setFlux(engineOutput);
// attach child geometry to be tranformed by the FCS
myFCS->addChild(myGeode);
...
// compute the data in the source pfFluxes to the engine
float *current = (float *)myData1->getWritableData();
... // compute data
myData1->writeComplete();

007-1680-080

533

Chapter 17

17. Active Surface Definition

Active Surface Definition (ASD) is a powerful real-time surface meshing and morphing
library. It enables you to roam surfaces that are too large to hold in system memory very
quickly. The surfaces, called meshes, are represented by triangles from more than one
LODs.
ASD is a library that handles real-time surface meshing and morphing in a
multiprocessing and multichannel environment. pfASD is an OpenGL Performer scene
graph node that allows you to place ASD information in a scene graph.
This chapter describes how to create and use ASD.

Overview
In the past, OpenGL Performer applications have dealt with large surfaces in two ways:
•

Level of detail (LOD), where the whole surface is one LOD.

•

Patches, where the surface is broken into geometrical subunits, each of which can be
at a different LOD level.

Each of these approaches has its disadvantages:
•

When the entire surface is one LOD, with large surfaces, memory limitations often
require such a high LOD that the resolution is poor.

•

Patches can only morph between adjacent LODs, which is insufficient in large
surfaces. The result is visible borders between LODs.

Active Surface Definition (ASD) is designed to solve these problems. ASD is a powerful
real-time surface meshing and morphing library characterized by the following features:

007-1680-080

535

17: Active Surface Definition

Transitions between different levels of detail (LOD) appear smooth and void of
spatial or temporal artifacts. Vertex position, normal, color, and texture coordinates
can be morphed.

2. LODs can be generated using simple adaptations of well-known,
non-uniform-tessellation surface subdivision algorithms.
3. ASD is real-time; in multiprocessor mode, it can sustain 60HZ operation while
displaying complex surfaces.
4. Nearly-coplanar objects, such as road surfaces, are accurately represented in the
non-uniform tessellation of each LOD using local textures.
5. Objects, such as buildings, may be modeled and rendered using alternate
algorithms, yet remain attached to the surface.
6. Triangles substantially outside the viewing frustum are culled from rendering.
7. The evaluation function that specifies the morphing factor for each geometry is
specified at run-time; this allows traversals to be optimized for different
applications and data sets.
8. Huge surfaces, such as one-meter data of the entire United State, are supported
using run-time paging from disk memory.
The pfASD approach uses a modeling surface that is a single, connected surface rather
than a collection of patches.
An ASD surface contains several hierarchical level-of-detail (LOD) meshes, where one
level encapsulates a coarser LOD than the next. When rendering an ASD surface, an
evaluation function selects polygons from the appropriate LODs and constructs a valid
meshing to best approximate the real surface on the screen.
Unlike existing LOD schemes, pfASD selects triangles from many different LODs and
combines them into a final surface. This feature lets a fly-through over a surface use
polygons from higher LODs for drawing nearby portions of the surface in combination
with polygons from low LODs that represent distant portions of the surface.
To review an ASD application, see perf/test/simpleASD/simpleASD.C.

536

007-1680-080

Using ASD

Using ASD
Modern computer graphics systems render objects represented by faceted surfaces
comprised of triangles. While rendering performance is increasing steadily, it remains
necessary to limit the number of triangles that are rendered if a real-time rendering rate
of 60 frames per second or greater is to be maintained.
Triangles may be omitted with no loss of image quality if they represent objects that are
not in the field of view (view frustum culling) or if they represent objects whose visibility
is occluded by other, nearer objects.
The number of rendered triangles can also be dramatically reduced with little or no loss
of image quality if multiple representations of each object are maintained, and the
representation used has as little detail as is necessary for displaying a good-quality
image. This technique and a body of others related to it are collectively referred to as level
of detail (LOD) reduction, and the multiple-object representations are known as LODs.

LOD Reduction
LOD reduction is not a new idea, having been practiced in the field of real-time image
generation for at least two decades. Early implementations eliminated the sudden,
visible transition from one level of detail to another by alpha blending two models over
a period of several frame times. More recently the fade transition has been replaced by a
geometric morph solution, which eliminates the need to temporarily render additional
triangles and interacts more nicely with depth buffer hidden surface elimination.
When the object being rendered is very large, there is no single level of detail that is
optimal for the entire object. Instead, it is desirable to render the object with a smooth
variation of level of detail; this allows nearer portions to be represented accurately, and
farther portions to be represented efficiently.
The need for spatially varying level of detail is particularly acute when rendering terrain,
which may range from several feet to hundreds of miles from the viewpoint. Early
attempts to support multiple LOD terrain modeled the surface as separate tiles, each of
which was rendered at a single level of detail. This approach results in physical
discontinuities between neighboring tiles rendered with different levels of detail, which
are visible as obvious ``walls.’’

007-1680-080

537

17: Active Surface Definition

Triangulated Irregular Networks

Triangulated irregular networks (TINs) can approximate a surface with fewer polygons
than other uniformly-grided representations. However, TIN models are hard to create
dynamically at interactive rates. Regular data, on the other hand, allows easier
construction of LODs at interactive rates. We propose a new terrain framework, Active
Surface Definition (ASD), that combines the advantages of both regular and irregular
networks. The terrain database is pre-computed and stored in a hierarchical structure of
efficient, irregular triangulations. LOD reduction is performed on the data structure at
real-time frame rates, varying levels of detail both spatially and temporally.

Hierarchical Structure
Active Surface Definition defines a hierarchical structure that organizes all the LODs of
a terrain. At run-time ASD traverses the structure, selecting triangles from different
LODs to render the portion of the terrain that is within the culling frustum. Triangles are
rendered either at pre-stored locations when they are in a particular fully-morphed
portion of a LOD range or at computed morphed locations when they are in the
morphing transition portions of the LOD range, as shown in Figure 17-1.

538

007-1680-080

Using ASD

D
LO

Morphing zones

Figure 17-1

Morphing Range Between LODs

Figure 17-1 shows the complete object represented by an LOD. ASD is often used for very
large terrains, like a map of the United States, where the entire terrain cannot fit into a
single LOD range. In that case, parts of the large terrain would be represented by
different LODs, as shown in Figure 17-2.

007-1680-080

539

17: Active Surface Definition

Figure 17-2

Large Geometry

In Figure 17-2, different parts of the terrain are displayed at different levels of resolution.

ASD Solution Flow Chart
To visualize large terrains using ASD, you use the following steps, which are portrayed
in Figure 17-3:
1.

Collect raw, elevation data.

2. Use a modeler to create ASD structures from the raw elevation data. There are
various triangle tessellation algorithms that can be used to model the terrain into
TINs. Adaptive tessellation and accurate representation and minimum visual
distraction between LODs are potentially all supported by pfASD structures.
3. Place the ASD structures into a scene graph by attaching them to pfASD nodes.

540

007-1680-080

A Very Simple ASD

4. Design an evaluation function that shows when a particular triangle on a particular
LOD level in ASD structure should be selected.
5. OpenGL Performer traverses and evaluates pfASD structures at run-time to create
an active mesh that is used for rendering.
Raw elevation data

Modeler

pfASD structures
Performer

Evaluation
function

Figure 17-3

Active mesh

ASD Information Flow

A Very Simple ASD
To demonstrate the ASD concept, we shall now build a very simple ASD surface. The
surface has two representations, which corresponds to two levels of detail (LOD). The
coarser level, LOD 0, has only a single triangle: T0. The finer level, LOD 1, has four
triangles: T1, T2, T3, T4, as shown in Figure 17-4.

007-1680-080

541

17: Active Surface Definition

V1
T0
LOD 0
V0

__
d2

__
d0

__
d1

V1
M0
T4

LOD 1
T3
M2

Figure 17-4

A Very Simple pfASD

pfASD requires that we specify a morphing relationship for the vertices in the two levels
of detail. This relationship guides ASD to choose triangles from the appropriate LODs
and place vertices in the correct positions.

Morphing Vector
Each triangle in LOD 0 can have up to four replacement triangles in a higher LOD. pfASD
imposes the limitation that the replacement of T0 must include the vertices V0, V1, V2,
and may include three more vertices. In our case, the replacement of T0 includes the
vertices V0, V1, V2, M0, M1, M2. We express the morph behavior between two levels by
describing the morph behavior of the triangle vertices.
Since V0, V1, and V2 remain unchanged, we need only specify the morphing behavior,
or the morphing vector, M0, M1, and M2. The morphing vector of a vertex is a vector
connecting the vertex to an edge of the lower resolution triangle. We specify morphing
vectors d0, d1, and d2 for the vertices M0, M1,and M2, respectively. The triangles in
LOD1 can now morph smoothly by morphing the vertex M0, M1, and M2 along their
morphing vectors d0, d1, and d2.

542

007-1680-080

ASD Elements

We call the edge position (M1+d1) unmorphed and, the M1 position fully morphed. The
unmorphed position is the reference position of a vertex because it is a reference to show
how a vertex should be coplanar with the triangle in the lower LOD when we replace the
lower LOD triangle with a higher LOD triangles, while providing the least amount of
visual distraction. The fully-morphed position is the final position of a vertex; that
position should be chosen from the original raw data by various triangulation
algorithms. It is recommended that the reference position be defined after the final
position is chosen to most accurately design the adaptive tessellation of a terrain. pfASD
handles the morphing of other attributes, such as normals, colors, and texture
coordinates as well.
This completes the construction of a very simple pfASD. When flying over this model—
for example, from V0 towards M1—pfASD shall morph M0 and M2 first and continue to
morph M1 as we get closer to the edge [V1,V2]. In other words, pfASD will morph closer
triangles into their higher LOD first.

A Very Complex ASD
To view a complex ASD, run perfly with the yosemite data supplied in the images.
Observe that the entire, visible terrain is represented by more than one LOD.

ASD Elements
pfASDs create active meshes to accurately and efficiently render terrains. A mesh is a
surface tessellated into triangles, as shown in Figure 17-2. Each frame contains an active
mesh. ASD reduces visual discontinuities between meshes by morphing geometry.
pfASD is defined by the following elements:

007-1680-080

•

vertices

•

triangles

•

evaluation function

543

17: Active Surface Definition

Vertices
The triangles of all LODs share a single pool of vertices: those vertices that define the
triangles of the highest LOD. A few vertices exist in all the LODs. These vertices define
the triangles in LOD0 and are typically chosen as corners of the terrain or other
significant features. Each of the remaining vertices exists in a contiguous set of LODs,
including the highest LOD. In order to facilitate the gradual introduction of a vertex into
the terrain, the actual position of each vertex is supplemented by a reference position. If
i is the lowest LOD that includes a vertex, then that vertex is located at its actual position
in LODs i through (n-1) and is morphed between its reference position and its actual
position during the morphing zone of LODi, which is adjacent to LOD(i-1) range.
Note: ASD could be extended to support multiple reference positions per vertex; this
allows vertices to morph between multiple LODs. This extension may be supplied in the
near future depending on demand.
A vertex is represented by its final position (pfVec3) and a morphing vector (pfVec3),
which represents the difference between the final position and the reference position. The
current position of a vertex is computed in this manner:
Vvertex = V0 + mVd

V0 is the final position of a vertex, m is the current morphing weight, and Vd is the
morphing vector. Notice that the reference position of a vertex is always chosen to lie on
an edge of a triangle from the lower LOD, though not necessarily at the centers of these
edges. This is to ensure that the replacement of the lower LOD by the higher LOD
happens when they are at coplanar positions; this eliminates sudden popping artifacts.
Attributes are represented by final position as well, as shown in Figure 17-5.

544

007-1680-080

ASD Elements

Final positions

Reference positions

Figure 17-5

Reference Positions

Normals, colors, and texture coordinates can all be linearly interpolated, as specified.
A vertex always morphs along the morphing vector, vd. When m = 1, the vertex is at a
no-morph position, which means it most likely is not part of the active mesh, because it
is on an edge of the parent triangle.
Triangles
A triangle exists in only one particular LOD. If the triangle is too coarse to accurately
represent the current terrain, the triangle is removed from the active mesh. The position
of a triangle is determined by the positions of its vertices, which may be morphed along
the morphing vector.
In the pfASD structure, each triangle node, called an pfASDFace, can have up to four
children. A sequence of frames may render the meshes shown in Figure 17-6.

Figure 17-6

007-1680-080

Triangulated Image

545

17: Active Surface Definition

As the distance between a vertex, P, and a viewer changes through the morphing range,
the side of the triangle associated with P is replaced by two edges from the next LOD.
Figure 17-7 shows how the next LOD triangles replace the current LOD triangles.

4
3

LOD1

Figure 17-7

LOD2

LOD1 Replaced by LOD2

Figure 17-7 shows the shapes of the triangles in two adjacent LODs.
Notice that only one side of the triangle on the left is changed; in this case, the original
triangle is now replaced by two triangles. The triangle that was on the right is now
completely gone; in its place are four triangles, as labeled. One triangle can be replaced
by up to four child triangles, depending on whether one, two, or three sides of the
triangles are replaced by two lines.
Sides shared by two triangles are evaluated only once. Morphing the shared side of one
triangle necessarily morphs it to the same position in the neighboring triangle.

Evaluation Function
The evaluation function determines the morph weight of every vertex. This function
returns a floating point value between 0.0 and 1.0 where 1.0 means the vertex is not
morphed and therefore not active. Since it is not active, it is not in the current mesh; its
geometry is represented by coarser LOD triangles. 0.0 means that the triangle is
completely morphed. In this case, the vertices are in their final positions in the active
mesh. Any number between 0.0 and 1.0 signifies a morphed position. OpenGL Performer
constructs a smooth mesh, based on the results from the evaluation function.
You can customize an evaluation function so that it applies to your application. For
example, instead of distance, your evaluation function might be based on altitude.

546

007-1680-080

Data Structures

For more information about the evaluation function, see “Default Evaluation Function”
on page 556.

Data Structures
Raw geometry data is converted into a tree-like, hierarchical structure of data, as shown
in Figure 17-8.

LOD0

LOD1

Figure 17-8

Data Structures

In Figure 17-8. each horizontal layer of triangles is equivalent to a single LOD mesh. The
nodes in the tree structure represent triangles in LOD layers of the terrain. Each triangle
contains three end vertices and up to three reference vertices. These vertices are indices
into the vertex and attribute arrays.
A vertex may be shared by multiple triangles within a single LOD and by many more
triangles throughout the data structure. A single position pair suffices for all the triangles
that share a vertex, but each triangle may specify its own vertex attribute pairs. This is
the reason that the vertex and attribute arrays are represented separately.
In Figure 17-8, the nodes in the top, root level represent the two LOD0 triangles shown
in Figure 17-6. The six triangles from LOD1, labeled C1 through C6, correspond to nodes
from the second level in the tree.
OpenGL Performer uses three data structures to encapsulate the triangle mesh
information:

007-1680-080

547

17: Active Surface Definition

•

Triangle—Encapsulates information about the sides, vertices, attributes, reference
points, parent, and children of a triangle.

•

Attributes—Encapsulates information about the attributes of each vertex, including
the normal, color, and texture coordinates.

•

Vertex—Encapsulates information about a morphing vector and the coordinates of
its final position.

Figure 17-9 shows the data structures, their fields, and their relationships.

pfs

Fac

fac

eID

lev

e
pfT

ver

rra

rt
Ve

tID

ver

t (3

)

attr

tDi

sat

orh

chi

ver
ghb

Figure 17-9

)

spl

it (3

ver
nei

tsid

(3)

D
fAS

Att

(3)
al
rm if (3)
o
n
alD
rm
no
(4)
lor
4)
o
C
if (
rD
o
l
Co
2)
d(
oo f (2)
c
T
i
dD
oo
Tc

t (3

ood

)

ld (

ASD Data Structures

The following sections describe these data structures in detail.

548

007-1680-080

Data Structures

Triangle Data Structure
The triangle data structure in PFASD contains information about the triangles in the
triangle mesh.
typedef struct pfASDFace
{
int level;
int tsid;
int vert[3];
int attr[3]
int refvert[3];
int refattr[3];
int child[4];
int refvert[3];
ushort gstateid;
ushort mask;
} pfASDFace;

Table 17-1 describes the fields in the triangle data structure.
Table 17-1

007-1680-080

Fields in the Triangle Data Structure

Field

Definition

level

Triangle’s LOD level.

tsid

Triangle strip ID.

vert[3]

Coordinates of the triangle’s three vertices.

refvert[3]

Coordinates of the triangle’s (up to) three reference points.

attr[3]

Pointer to attribute values for each of the triangle’s three vertices.

refattr[3]

Pointer to attribute values for each of the triangle’s (up to) three reference
points.

child[4]

Triangle IDs of the triangle’s (up to) four child triangles.

gstateid

GeoState ID.

mask

Specifies whether or not to render the triangle.

549

17: Active Surface Definition

Triangle IDs and LOD Levels

Each triangle in the database has an ID. Since one triangle can be replaced by as many as
four triangles, child[4] is an array of child triangle IDs. The child IDs may be listed in any
order in the array. Do not list the same child more than once in the child array.
If a triangle does not have four children, as is the case of the triangle on the left in
Figure 17-7, enter the token PFASD_NIL_ID in the child[4] array where normally you
would enter the IDs of the child triangles.
An ASD node may very likely have more than one texture in it; so, it contains more than
one pfGeoState; one for each texture. gstateid specifies the pfGeoState object the triangle
face uses.
There are times when you might like to have a triangle face specified in the structure but
not draw it. For example, you need the triangles of the ground under an airport so that
you can place the airport on it. However, since the triangles are covered by the airport,
you do not want to render them. mask allows you to prevent the rendering of the triangle
face specified in the structure.
As shown in Figure 17-8, each triangle is resident in a particular LOD level. level defines
the LOD of the triangle. If the viewer is moving within one LOD, chances are they need
to see other triangles at the same LOD level.
Discontinuous, Neighboring LODs

There are many cases in which you can have neighboring triangles with discontinuous
LOD levels; for example, a high-resolution insert of a scene might be LOD level 5 and
surrounded by triangles of LOD 1.
When neighboring triangles have discontinuous LOD levels, the lower-resolution
triangles must have the following, as shown in Figure 17-10:

550

•

reference vertices on all triangle edges

•

PFASD_NIL_ID entered in the fields for all four of its children

007-1680-080

Data Structures

Reference vertices defined

Discontinuous, Neighboring LODs

Figure 17-10

Triangle Strips

You can increase the rendering speed of your application by using triangle strips.
Triangle strips are groups of contiguous triangles that together form a strip, as shown in
Figure 17-11.

11
10
7

6
9
8
5
4

Figure 17-11

Triangle Mesh

Triangle strips improve rendering performance because, after the first triangle in the
strip, each additional triangle can be added by defining only one additional vertex.

007-1680-080

551

17: Active Surface Definition

The pfASD does not generate triangle strips dynamically. Triangle strips are
pre-generated at modeling time and represented using tsid in the pfASDFace structure.
Only triangles from the same ASD face tree level can be in the same triangle strip. The
tsid numbers are sorted at run-time to connect as many triangles from the same triangle
strips together as possible. Triangles with consecutive tsid values are placed in one
triangle strip.
The pfASD does the following:
1.

Evaluates the LOD structure.

2. Picks the triangles at the desired LODs.
3. Culls the triangles to the visible frustum.
4. Sorts the triangles by their tsid.
5. Generates triangle strips of triangles with consecutive tsid values.
Figure 17-12 shows an example of a set of triangles with consecutive tsid values.

0
2

tsid=6
tsid=8

tsid=10

tsid=3
tsid=7
1
1 1
2

0
0

tsid=9
2

tsid=2

tsid=11
1
2

Figure 17-12

Using the tsid Field

pfASD generates a triangle strip from the vertices marked 0, 1, and 2 in the triangle with
tsid=2, followed by the vertex marked 2 in the triangle with tsid=3. The triangle with
tsid=6 marks the beginning of the next triangle strip, vertices marked 0, 1, and 2 in
triangle with tsid=6, followed by vertices marked 2 from triangles with tsid 7, 8, 9, 10, and
11. It is very important to start a triangle strip on an even triangle strip ID. For example,
triangles in the first triangle strip can have IDs 0, 1, 2...; but they cannot have 1, 2, 3,...
Vertices should be listed in pfASDFace in counter-clockwise order with the last vertex to
be the vertex in the triangle strip.

552

007-1680-080

Data Structures

At the termination of a triangle strip, there must be a non-consecutive tsid in the next
triangle, as shown in Figure 17-12. Assigning the numbers 3 and 6 forces these triangles
into separate triangle strips. Assigning all triangles to the same tsid would separate all of
them into strips with only one triangle.
Note: pfASD sorts all the visible triangles by tsid. tsid values, therefore, must not overlap
in different areas of the surface. pfASD does not check that triangles with consecutive tsid
values are indeed neighboring.

Vertex and Reference Point Arrays

The vertex and reference point arrays, vert[3] and refvert[3], each contain three
pfASDVerts, which are the coordinates of the vertices and reference points. For more
information on pfASDVert, see “Vertex Data Structure” on page 555.
The order of the vertices and the reference points in the arrays must be
counter-clockwise, as shown in Figure 17-13.
1
C
B
0
2

Figure 17-13

Vertex and Reference Point Arrays, Counter-Clockwise Ordering

Any of the vertices can be the first in the array. The first reference point in the array,
however, must be the one on the side between the first and second vertex points; for
example, the reference point on the hypotenuse of the triangle in Figure 17-13 should not
be entered as the first reference point in the refvert{3] array because the bottom leg of the
triangle contains the first two vertex points entered into the vert[3] array.
If a triangle does not have three reference points, as is the case of the triangle on the left
in Figure 17-5 (where only the hypotenuse has a reference point), enter the token
PFASD_NIL_ID in the refvert{3] array where normally you would enter the indices of the
reference points.

007-1680-080

553

17: Active Surface Definition

Attributes

Each vertex can have its own set of attributes. The attribute format is described in
“Attribute Data Array” on page 554. The arrays, attr[3] and refattr[3], point to three
vertex attribute array entries and three reference point attribute array entries.
If a vertex or reference point is not assigned an attribute, enter the token PFASD_NIL_ID
in the attr[3] or refattr[3] arrays where normally you would enter the pointer to the
attribute array.

Attribute Data Array
Each vertex and reference point can be assigned an attribute. An attribute is an
interleaved list of floats that consists of normal, color, and texture definitions, as follows:
float
float
float
float
float
float

normal[3];
normalDif[3];
Color[4];
ColorDif[4];
Tcood [2];
TcoodDif [2]

If any attribute is not defined, it should not be represented in the array, in which case the
stride is shorter.
Each attribute has two listings: the values of the attributes at the final positions and how
much they change during morphing. For example, the normal[3] array specifies the
vector that is normal to the vertex at the final position while the normalDif[3] array
specifies the difference between the normal and its final position and the normal at its
reference point.
Color{4] and ColorDif[4] similarly specify the RGBA values at a vertex and the color
morphing vector.
The texture coordinates, Tcood [2] and TcoodDif [2], specify the texture coordinates of
vertices at their final position and the texture coordinates of the morphing vector.
Setting the Attributes

The attribute array of floats can contain normal, color, and texture coordinate
information, or it can contain any combination of attributes. You specify which of the

554

007-1680-080

Data Structures

attributes are in the attribute array of floats by setting the pfASDAttr() mask. To specify
that you are including some or all of the attribute values in the attribute array, you use
pfASDAttr() with one or more of the following tokens ORed together:
•

PFASD_NORMALS

•

PFASD_COLORS

•

PFASD_TCOORDS

For example, to specify normal and color attributes only, use the following statement:
pfASDAttr(PFASD_NORMAL | PFASD_COLOR);

If certain attributes are not specified in pfASDAttr, they should not be included in the
attribute array. pfASD only takes one set of per-vertex-per-face attributes and one set of
overall attributes.
Global Attributes

If you want to use the same attribute value for all normals, colors, or texture coordinates,
you can specify a global attribute value rather than making the same entry for each
normal, color, or texture coordinate attribute. To specify a global attribute, use the
following statement:
pfASDAttr(..., PFASD_OVERALL_ATTR, attr);

attr, an array of floats, holds a set of attributes and their morphing vectors. You can mix
a global attribute with a different attribute that is per vector.

Vertex Data Structure
The vertex data structure, defined as follows, holds the coordinate information of
vertices and their vertex vectors.
typedef struct pfASDVert
{
pfVec3 v0, vd;
int neighborid[2];
int vertid;
} pfASDVert;

007-1680-080

555

17: Active Surface Definition

Each vertex has an ID, vertid; a set of coordinates, vo; and a vertex vector, vd. The vertex
vector is the vector stretching from the final vertex toward the reference point.
In a triangle strip, a line segment is commonly shared by two adjoining triangles, as
shown in Figure 17-14.

Neighborhood (1)

Neighborhood (0)

Figure 17-14

Vertex Neighborhoods

The new vertex is shared by the adjoining triangles. Those triangles are referred to as the
neighborhoods of the vertex. The neighborhood array contains the triangle IDs of the
adjoining triangles in the previous LOD.
If a side is not shared by two triangles, then one of the neighborhood array values should
be PFASD_NIL_ID.

Default Evaluation Function
The default evaluation function, which returns a value between 0.0 and 1.0, is based on
the distance between the vertex and the eyepoint: the farther away a vertex is, the lower
its resolution.
To use the default evaluation function, you must fill in the pfASDLODRange structure
for each LOD.

556

007-1680-080

Data Structures

typedef struct pfASDLODRange
{
float switchin;
float morph;
} pfASDLODRange;

LOD[i].switchin is the far edge of the LOD[i]. Take any vertex of a triangle in LOD[i] and
compute the distance between the vertex and eyepoint. If this distance is more than
LOD[i].switchin, the morphing weight of this vertex is 1.0 (NO_MORPH). If it is less, the
morphing weight is calculated against the morphing zone.
LOD[i].morph is the length of the morphing zone from the LOD[i].switchin far edge. The
morph weight is (1-(LOD[i].switchin-dist)/LOD[i].morph). If the distance is less than
LOD[i].switchin - LOD[i].morph, the morph weight is 0.0 (COMPLETE_MORPH).
Overriding the Default Evaluation Function

The evaluation function can be based on many things beside distance. For example, you
might implement the evaluation function based on the following:
•

Line of sight—Geometries directly in front of the camera have the highest resolution
while geometries more peripheral have a lower resolution.

•

Direction—High-resolution LODs appear only in one direction; for example, given
the fuzzy nature of clouds, when a pilot looks straight ahead, the clouds do not
need to be rendered with great detail. When the pilot looks down, however, you
might like them to see the terrain in great detail.

•

True size of projected triangle—Triangles are replaced when they physically reach a
specified size on the projector screen.

•

Event driven—An event triggers a change of LODs, for example, when a bomb
explodes.

•

Shape—Specific shapes in the scene, perhaps a tank, could be rendered in higher
resolution than the surrounding countryside.

None of these evaluation criteria can use the default evaluation function. Instead, you
must write your own evaluation function and register it as a callback function with
pfASD using the following method:
extern void pfASDEvalFunc(pfASD* _asd, pfASDEvalFuncType _eval);

007-1680-080

557

17: Active Surface Definition

The evaluation function you provide, eval, must return a float between 0.0 and 1.0 and
conform to a fixed format. pfASDEvalFuncType takes the argument vertid, which is the
index of the vertex from the next LOD whose reference position is on this edge. Refer to
asdfly/pfuEvalFunc.c for examples.

pfASD Queries
pfASD supplies a query mechanism for two kinds of objects:
•

vertices with a down vector
A vertex query answers the question: Where does the ray through the vertex in the
given down direction hit the surface?

•

triangles with a projection vector and a down vector
A triangle query answers the question: given a triangle, how does it project onto the
current morphing pfASD? Or, given a triangle, supply a list of triangles describing
the projection of this triangle on the surface.

The pfASD has a mechanism for manipulating arrays of vertices and triangles and for
reporting query results of surface geometries. Query results are kept in a pfFlux buffer.
The application may read this buffer directly or connect it to a pfEngine for processing.
These query mechanisms must run in sync with the evaluation function. The query
results must be used in the same frame that the morphing status of a pfASD becomes
active. To synchronize the behavior of pfASD geometry and query points, you need to
set the correct pfFlux flags, as described in “Aligning Light Points Above a pfASD
Surface Example” on page 564.
Before describing the technical details of setting up a query array, the following section
describes how to use these arrays.

Aligning an Object to the Surface
In the visual simulation arena, you often place objects on surfaces. Since the pfASD
surfaces morph, the objects on top must move to remain on the surface. Some objects
require only position changes. Others require angle changes as well. A building on top
of a mountain side, for example, requires only a position change to match the current

558

007-1680-080

pfASD Queries

altitude of the surface under it. A car standing on the same mountain side, however,
requires changing both its position and its angle in order to remain aligned to the surface.
In the case of the building, it is enough to query the intersection between a query ray and
the surface. In the case of the car, however, you must also calculate the surface normal at
the intersection point.
Casting a Shadow

We often wish to cast the shadow of objects on the surface underneath them. Since the
pfASD surface is morphing, we must morph the shadows accordingly. Given a triangle
and a projection direction, we want to project the triangle on the surface and tessellate its
projection so that we get a 3D representation of the shadow. We can later use these
shadow triangles as decals to paint the shadow on the surface.
Generating Decals

Many times you wish to draw decals on a surface. In the context of surfaces, you may
wish to add a road as a decal on the morphing surface. To make the road touch the
surface, you must tessellate the polygons of the road so they conform to the shape of the
surface. To tessellate the road, you project a triangle downwards and then tessellate the
result.

Adding a Query Array
To add an array to the pool of queries, you use the following:
•

pfASD::addQueryArray() to add an array of vertices.

•

pfASD::addQueryGeoSets() to add an array of triangles.

Both methods require a pfFlux as input. After completing the evaluation of a pfASD
frame, pfASD fills the specified pfFlux objects with query results for all the visible query
arrays. The query results become relevant at the next pfFrame along with the newly
evaluated pfASD geometry.
To make sure the query results are used only when their corresponding geometry is
active, add the corresponding pfFlux and the pfASD node to the same sync group using
pfASD::setSyncGroup() and pfFlux::setSyncGroup().

007-1680-080

559

17: Active Surface Definition

Using ASD for Multiple Channels
A channel is a view of a scene. There are three approaches to linking multiple channels:
•

Each channel uses the same database, the same viewpoint, the same evaluation
function, and the same LODs, but the viewing angle is different for each channel.
The effect is multiple channels that together create a continuous view like that of a
horizon. The views of the scenes in each channel move together. For more
information on this approach to multiple channels, see “Controlling Video
Displays” on page 482.

•

Each channel uses the same database, a different viewpoint, the same evaluation
function, and potentially different LODs.
The effect is two or more views of the same scene; for example, one view might look
along the terrain and another view might look down at the same terrain from
above. This approach uses the pfASD API for creating multiple channels, as
described in “Connecting Channels” on page 560.

•

Each channel uses the same database, a different viewpoint, different evaluation
functions, and potentially different LODs.
The effect is two or more views of potentially unrelated parts of a scene; for
example, if the database is the earth, one view might be of someone traveling
around the north pole and another view might be of someone traveling around the
south pole.
Because the only thing common to the two views is the database, you need to create
two pfASD objects, one to handle each view.

The following section explains how to do the second approach to connecting channels.

Connecting Channels
When you have two channels that share the same database and evaluation function and
differ only in that they have different viewing angles of the same part of a scene, you can
save processing time and save memory by attaching the two channels using
pfChanASDattach(); pfChanASDdetach() detaches two channels.
By making the two channels part of the same pfASD, processing time is improved
because only one evaluation function is computed. Having one pfASD instead of two
saves memory because two pfASDs requires maintaining two sets of the same scene.

560

007-1680-080

Combining pfClipTexture and pfASD

Combining pfClipTexture and pfASD
A pfClipTexture is able to manage very large bodies of data, presenting only the part that
is in view of the user. pfASD modifies the pfClipTexture such that the different parts of
the pfClipTexture are presented using the correct LOD level.
To use a pfClipTexture as part of a pfASD, use the following procedure:
1.

Put the pfClipTexture into a pfGeoState.

2. Set the pfGeoState to a pfASD.
pfASD accepts a list of pfGeoStates.
3. Define the center of the pfClipTexture in one of two ways:
•

Attach a pfuClipCenterNode to a pfASD, as shown in the .im loader.
Use pfuTexGenClipCenterNode and convert the eyepoint through texgen into
texture space as the clipcenter. This technique should be used when cliptexture
is applied using texgen.
Set the proxy box equal to the rough boundary of the pfASD database. The
boundary of the pfASD can be queried by calling pfGetASDBBox(asd, box).

•

Use a PRE callback function with pfASD to set the clip center yourself. To see an
example, refer to terrain.c.

4. If the clip texture is virtual, pfASD automatically sorts geometry into concentric
cliprings to support the appropriate cliptexture limit and offset. Set the
environment variable, PFASD_CLIPRINGS to activate this feature. See libpfdem
and libpfevt for examples.

ASD Evaluation Function Timing
The pfASD evaluation function commonly runs as a separate, asynchronous process.
This process does not necessarily finish its evaluation when the App-Cull-Draw frame
finishes. The new pfASD geometry becomes active (visible) at the end of the pfFrame that
follows the completion of the pfASD evaluation.
For example, if Draw frames 0, 1, 2, 3, 4... end at times 100, 200, 300, 400, 500... and the
pfASD evaluation finishes at time 320, the new geometry is introduced into the scene
graph at time 400, as shown in Figure 17-15.

007-1680-080

561

17: Active Surface Definition

Draw frame:
ASD
Geometry
ASD
Evaluation
Frame

asd0

asd1

Time
100

200

300

400

500

600

The pfASD evaluation process
completes at time 320 but the
new pfASD geometry enters
the scene graph at time 400

Figure 17-15

pfASD Evaluation Process

In Figure 17-15, it is important to align objects to the old geometry in Draw frame 2 and
to align to the new geometry in Draw frame 3.

Query Results
The query results must affect the aligned geometry in the same frame that the pfASD
geometry becomes active. To achieve this, you must connect both the pfASD node and
Matrix-Flux to the same sync group. In this way, when pfASD changes the active surface
geometry, it also activates the newly calculated query results.
In general, you should move as much processing to the pfASD process as possible. In
Figure 17-15, all the query/alignment operations take place in the pfASD process. Since
the entire evaluation is triggered by the query array writing to Result Flux, the
generation of the new pfFCS matrix in Matrix Flux also takes place in the pfASD process.
The only change to the time critical APP-CULL-DRAW sequence is a single
pfFlux::getCurData() to get the updated pfFCS matrix.

Aligning a Geometry With a pfASD Surface Example
To align some geometry to a pfASD surface using query points, use the following
procedure:

562

007-1680-080

ASD Evaluation Function Timing

Pick an anchor point where you want the target geometry to go.

2. Generate a query array containing the anchor point and add it to pfASD.
3. Request pfASD to store the query results in the pfFlux marked Result Flux.
4. Generate an additional pfFlux and connect it to the pfFCS node matrix controlling
the target geometry.
Figure 17-16 diagrams this procedure.

C
pfF

e
od

t
rge
Ta etry
om
ge

Figure 17-16

trix

S
get

rDa

SD 1
pfAroup=
G
ync
ery
Qu ys
a
r
ar

get

itab

pfE
n
(AL gine
IGN
)

le
lu
tF
sul H)
e
R US
(P

x
Flu
trix up=1
a
o
M Gr
nc
Sy

Example Setup for Geometry Alignment

At run-time pfASD does the following:
1.

pfASD starts its frame by evaluating the ASD surface and generating optimized
geometry for it.

2. pfASD calculates the query array values.
3. pfASD writes the results into the pfFlux marked Result Flux.
4. The write operation triggers a calculation of the pfEngine because the Result Flux is
a PUSH flux.
5. The pfEngine generates a transformation matrix and stores it in the pfFlux marked
Matrix Flux.

007-1680-080

563

17: Active Surface Definition

6. The pfFCS node controlling the transformation of the target geometry has a fluxed
matrix.
7. When the pfFCS node is traversed, it retrieves the newly generated matrix from
Matrix Flux and uses it to align its child geometry.
Note: If you run the pfASD evaluation function in a separate process, the evaluation of
the query arrays may happen asynchronously at some point during the OpenGL
Performer frame.

Aligning Light Points Above a pfASD Surface Example
The following example, diagramed in Figure 17-17 demonstrate how to link pfASD
queries to a pfEngine. For a GeoSet containing light point primitives, this code places all
the primitives at some offset above the surface.

asd

ge
att

flux

Figure 17-17
Example 17-1

pfE
n
(AL gine
IGN
)

u
res

lt_f

se
off

lux

t_a

rra

Aligning Light Points Above a pfASD Surface
Aligning Light Points Above a pfASD Surface

// Generate a flux for pfASD to store the query results.
results_flux = pfNewFlux(nofLightPoints *
sizeof(pfVec3),PFFLUX_DEFAULT_NUM_BUFFERS, pfGetSharedArena());
// Make so that writing into this flux will trigger evaluation of
// connected engines.
pfFluxMode(results_flux, PFFLUX_PUSH, PF_ON);

564

007-1680-080

ASD Evaluation Function Timing

// Generate final flux - this flux will contain the final aligned
// light points.
attr_flux = pfNewFlux(nofLightPoints * 3 * sizeof(float),
PFFLUX_DEFAULT_NUM_BUFFERS, pfGetSharedArena());
// Add flux to the same syncGroup as its aligning pfASD.
pfFluxSyncGroup(attr_flux, 1);
// Initalize the flux to the unaligned point positions.
// The engine will only modify a portion of the buffer (the Z
// values), so we must initialize the (X,Y) values now.
// Assume vertexList contains the original vertices.
FluxInitData(attr_flux, vertexList);
// Add the array of positions to pfASD. Request position-only
// queries.
query_id = pfASDAddQueryArray(asd_hook, vertexList, down,
nofLightPoints, PR_QUERY_POSITION, results_flux);
// Get maximum bounding box of the vertex array - this box
// contains all possible positions of the query array points.
pfASDGetQueryArrayPositionSpan(asd, query_id, &box);
// Generate a SUM engine to sum the query vertex results with a
// constant array.
sum_engine = pfNewEngine(PFENG_SUM, pfGetSharedArena());
pfEngineIterations(sum_engine, nofLightPoints, 1);
// Inform pfEngine of its two sources and one destination.
// offset_array should contain a constant offset for each light
// point. We request sum_engine to modify the Z coordinate of
// each light point vertex.
pfEngineSrc(sum_engine, PFENG_SUM_SRC(0), offset_array, NULL,
0, PF_Z, 3);
pfEngineSrc(sum_engine, PFENG_SUM_SRC(1), results, NULL,
0, PF_Z, 3);
pfEngineDst(sum_engine, attr_flux, NULL, PF_Z, 3);
// Generate GeoSet to hold the final aligned geometry.
gset = pfNewGSet(pfGetSharedArena());
// standard pfGeoSet initialization code
// (GeoState, color, NumPrims, etc) omitted.
pfGSetPrimType(gset, PFGS_POINTS);

007-1680-080

565

17: Active Surface Definition

// Add flux_attr as the COORD3 attribute of the geoset.
pfGSetAttr(gset, PFGS_COORD3, PFGS_PER_VERTEX,
(void *) attr_flux, NULL);
// Set the GeoSet bounding box staticly. This will help avoiding
// recalculation or mis-calculation of bounding boxes for culling.
pfGSetBBox(gset, &box, PFBOUND_STATIC);
// Setup range based evaluation. assume ‘center’ contains the
// center of box.
pfEngineEvaluationRange(sum_engine, center, 0.0, 40000.0);
pfEngineMode(sum_engine, PFENG_RANGE_CHECK, PF_ON);

Paging
Large scale surface simulations require large amounts of memory to store high resolution
surfaces. For example, the earth sampled from a height of 100m requires tens of gigabytes
of disk space. Such large amounts of memory cannot reside entirely in system memory.
Consequently, efficient database paging is essential in supporting a 30 Hz frame rate.
A common paging method is to page in tiles. Each area block contains all the LOD
information describing an area. The problem with this method of paging is that in a
surface that extends far into the distance, the large number of tiles visible to the viewer
consumes more memory than is available on a system.
The hierarchical structure of ASD provides a different paging method: LOD paging. Each
LOD is divided into a set of blocks; each block represents an area of the scene at a
particular resolution. These tiles are paged in independently. All tiles for a single LOD
are the same size; tiles for different LODs, however, can be different sizes.

Interest Area
In one LOD, the blocks most likely accessed, according to the view and position of the
viewer, are called the interest area of the LOD. It is these blocks that get paged in.
The interest area for an LOD whose resolution is based on range is bounded by the
maximum range value for that LOD. The interest area is smaller for higher resolution
LODs, larger for lower resolutions LODs. As a result, the memory requirements of ASD
are reduced because not all LODs in an area are paged in: triangles closer to the viewer

566

007-1680-080

Paging

are paged in at a higher resolution LOD than those triangles further from the viewer, as
shown in Figure 17-18.

Paging area

Interest area

Figure 17-18

Tiles at Different LODs

Each square in Figure 17-18 holds the same amount of information: the large pages hold
a large amount of low-resolution information while the small pages hold a small amount
of high-resolution information.
If the observer does not make discontinuous jumps in location between 2 frames, you use
an algorithm that anticipates what the viewer needs to see next based on the evaluation
function. For example, if the evaluation function is based on distance, you would page
blocks into memory in the direction you expect the viewer to go and release from
memory those pages the viewer is leaving.

Preprocessing for Paging
Preprocessing pages improves performance.
1.

Determine the appropriate tiles size for each LOD based on the evaluation function.

2. Assign triangles in each LOD into tiles and store each tile in a paging unit, such as a
file.

007-1680-080

567

17: Active Surface Definition

Assign a triangle to one tile only. Triangles with the same parent node must be
assigned to the same tile.
3. Generate a set of fast-paging files for a specific set of paging areas.
When the application is run, it should anticipate the pages the viewer will need and then
page them into memory from disk and preprocess them.
Order of Paging

Paging in lower LOD pages before higher LOD pages ensures that a page at some level
of LOD is always ready for the viewer. When you travel so fast through a surface that the
higher resolution LODs do not have time to load, the surface has a lower resolution,
which is what you would expect to see when traveling fast.

Multi-resolution Paging
pfASD supports multi-resolution paging. The format of a paging file, where all of the
geometries are in one tile, is described by the following structure:
int numfaces
int numverts
/* numfaces of the following */
int faceid1
/* structure of the face faceid1 */
pfASDFace face
/* structure of the face faceid2 */
int faceid2
pfASDFace face
...
/* numfaces of the following */
/* face bounding box of faceid */
pfBox box1
/* face bounding box of faceid2 */
pfBox box2
...
/* numverts of the following */
int vertid1
/* structure of vertex vertid1 */
pfASDVert vert
int vertid2
/* structure of vertex vertid2 */
pfASDVert vert

568

007-1680-080

Paging

...

To calculate the paging area, use the following code:
page[0] = (int)(lods[i].switchin/tilesize[0]) + lookahead[0];
page[1] = (int)(lods[i].switchin/tilesize[1]) + lookahead[1];

lods[i] is the pfASDLODRange of LOD[i]. tilesize is the size of the tile in LOD[i]. lookahead
is the number of extra tiles in one direction to page in to memory to overcome a paging
delay
To process the paging tiles into a fast paging format, use pfdProcessASDTiles() in
pfdProcASD.c. pfdProcessASDTiles() returns a new set of files as xxxx.asd. These
files are paged in real-time.
For an example of writing a file, see pfdWriteFile() in libpfdu/pfdBuildASD.c.

007-1680-080

569

Chapter 18

18. Light Points

OpenGL Performer provides sophisticated light point objects:
•

pfLPointState lights are emissive objects that do not illuminate their surroundings,
such as stars, beacons, strobes, runway edge, and taxiway lights.

•

pfCalligraphic light points are pfLPointState light objects with calligraphic
extensions; they are intensely bright light points that require special display
equipment.

Uses of Light Points
Light points are bright points of light that have the following characteristics:
•

intensity

•

directionality

•

attenuation shape

•

distance computation

•

attenuation through fog

•

size and fading

These attributes make light points excellent for use as stars, beacons, strobes, runway
edge and end illumination, and taxiway lights.
For example, three light points with specific directionality may be used to create a VASI
light system at an airport that appears as follows:

007-1680-080

•

Red when the pilot is above the landing glide path.

•

Green when the pilot is below the landing glide path.

•

White when the pilot is on the landing glide path, as shown in Figure 18-1.

571

18: Light Points

VASI
Red
White
Green

Figure 18-1

VASI Landing Light

Creating a Light Point
To create a light point, do the following::
1.

Create a pfGeoSet of points (PFGS_POINTS):
pfGeoSet *lpoint = pfNewGSet(arena);
pfGSetPrimType(lpoint, PFGS_POINTS);

2. Create a pfLPointState and define its mode and values:
pfLPState *lpstate = pfNewLPState(arena);
pfLPStateMode(lpstate, PFLPS_SIZE_MODE, PFLPS_SIZE_MODE_ON);
pfLPStateVal(lpstate, PFLPS_SIZE_MIN_PIXEL, 0.25f);
pfLPStateVal(lpstate, PFLPS_SIZE_ACTUAL, 0.07f);
pfLPStateVal(lpstate, PFLPS_SIZE_MAX_PIXEL, 4.0f);

3. Attach the pfLPointState to the pfGeoState associated with the pfGeoSet:
pfGStateMode( gstate, PFSTATE_ENLPOINTSTATE, PF_ON);
pfGStateAttr( gstate, PFSTATE_LPOINTSTATE, lpstate);
pfGSetGState( gset, gstate ) ;

This pfGeoSet must have a PFGS_COLOR4 attribute binding of
PFGS_PER_VERTEX.

572

007-1680-080

Setting the Behavior of Light Points

Setting the Behavior of Light Points
The pfLPointState attached to a pfGeoSet controls the behavior of all light points in the
pfGeoSet. For example, the color, position, and direction vector of each light point is set
by the color, vertex, and normal stored in the pfGeoSet.
You use pfLPStateMode, to enable or disable the behavior, and pfLPStateVal to set the
following pfLPointState values:
•

Intensity—setting the intensity of light from no attenuation, 1, to fully attenuated, 0.

•

Directionality—specifying the direction and the shape of the emanation of light.
The shapes are either a single or a pair of opposite-facing elliptical cones.
pfLPStateShape() specifies the shape of the emanation.

•

Fading—specifying how the light point fades when receding into the distance.
Fading is often more realistic than simply shrinking the point size to 0.

•

Fog punch-through—specifying how a light point is obscured by fog.

•

Size—setting the maximum and minimum size of the light point based on
perspective.

The following sections describe how to set these values.

Intensity
The intensity of the light points in a pfLPointState can be attenuated using
pfLPStateVal(), as follows:
pfLPStateVal(lpstate, PFLPS_INTENSITY, intensity)

The intensity value, intensity, must range between 0 and 1; 1, no attenuation, is the
default.
The intensity of each light point is defined by its four component colors.

007-1680-080

573

18: Light Points

Directionality
Each light point has a normal, which defines the direction of its emanation. By default,
all of the light points in a pfLPointState point in the same direction. That direction is
defined as the “front” of the light point.
If all light points in a pfGeoSet share the same normal, use the PFGS_OVERALL attribute
to optimize the light point computation.
If the light points are directional, the pfGeoSet attributes must specify normals.
Enabling Directionality

To enable or disable light point directionality, use the following method:
pfLPStateMode(lpstate, PFLPS_DIR_MODE, mode);

The mode value can be one of the following:
•

PFLPS_DIR_MODE_ON, the default, turns on directionality computation.

•

PFLPS_FOG_MODE_OFF disables directionality computation.

To specify whether or not the light point emanates towards the front and back, use the
following method:
pfLPStateMode(PFLPS_SHAPE_MODE, mode);

The mode value can be one of the following:
•

PFLPS_SHAPE_MODE_UNI, the default, makes the light point not visible from the
back side.

•

PFLPS_SHAPE_MODE_BI makes the light point bidirectional, the behavior is the
same for both sides.

•

PFLPS_SHAPE_MODE_BI_COLOR makes the light point bidirectional, but if seen
from the back side, the light change its color. This color is set using
pfLPStateBackColor().

Emanation Shape
The emanation shape is an elliptical cone, as shown in Figure 18-2.

574

007-1680-080

Setting the Behavior of Light Points

Normals

Eye point

Eye point angle
Light points

Figure 18-2

Attenuation Shape

Optionally, the intensity of light can fall off from the normal to the edge of the cone.
The direction of emanation is defined along the Y axis and the shape is defined in the
local coordinate system. The coordinate system is rotated so that the Y axis aligns with
the normal of each light point.
To specify the shape of emanation, use the following method:
pfLPStateShape(lpstate, horiz, vert, roll, falloff, ambient)

Shape

horiz and vert are total-angles (not angles to the normal) in degrees, which specify the
horizontal and vertical dimensions of the cone about the normal. The maximum value
for these angles is 180 degrees, which creates a non-directional light point. The default
values for horiz and vert is 90 degrees.
Tip: A symmetric cone (where horiz and vert are equal) is faster to compute than an
asymmetric cone.

007-1680-080

575

18: Light Points

Rotation

The cone is rotated by roll degrees through +Y. The default roll value is 0.
Falloff

When the vector from the light’s position to the eye point is outside the cone, the light
point’s intensity is ambient. The default ambient value is 0.
If the vector from the light’s position to the eye point is within the cone, the intensity of
the light point is based on the angle between the normal and that vector. The intensity
values range from 1.0, when the eye point lies on the normal, to ambient, at the edge of
the cone. How the light attenuates between the normal and the edge of the cone is
specified by falloff.
falloff is an exponent that modifies the light point’s intensity. A value of 0 indicates that
there is no falloff; a value of 1, the default, indicates a linear falloff; values greater than 1
indicate a progressively greater geometric falloff, as shown in Figure 18-3.
Light
intensity
falloff = 0

bie

Angle deviation
from normal

Figure 18-3

falloff > 1

falloff = 1

Am
nt

Light
intensity

of
ge
Ed e
con

bie

Angle deviation
from normal

o
ge
Ed e
n
o
c

Angle deviation
from normal

Attenuation of Light

Distance
The distance between the light point and the eye point is used when computing the light
point’s size and intensity through fog. Since these calculations are intensive, OpenGL

576

007-1680-080

Setting the Behavior of Light Points

Performer provides an approximation for the distance between the light point and the
eye point using the following method:
pfLPState(lpstate, PFLPS_RANGE_MODE, mode)

The mode value can be one of the following:
•

PFLPS_RANGE_MODE_TRUE approximates the distance using the depth (Z)
difference between the eye and the light point.

•

PFLPS_RANGE_MODE_DEPTH, the default, uses the real distance between the
eye and the light point.

Tip: The wider the field of vision (FOV), the less accurate is the approximation.

Attenuation through Fog
Light points are visible at greater distance than non-emissive polygons. For that reason,
the light points appear to punch through fog. To account for this higher visibility, you can
supply a punch-through value in the following method:
pfLPStateVal(PFLPS_FOG_SCALE, punch-through);

punch-through is a float that is multiplied times the distance. When punch-through is less
than 1.0, the product of the distance and the punch-through value is less than the distance.
The revised distance is used in calculating the apparent brightness of the light point;
when the punch-through is less than 1.0, the light point appears brighter than it otherwise
would. The default punch-through value is 0.25.
Enabling Punch-Through

To use punch-through, you must first enable it using the following method:
pfLPStateMode(lpstate, PFLPS_FOG_MODE, mode);

The mode value can be one of the following:

007-1680-080

•

PFLPS_FOG_MODE_ON, the default, enables fog punch-through computation.

•

PFLPS_FOG_MODE_OFF does not modify the distance before fog is applied.

577

18: Light Points

Size
Light points can exhibit perspective behavior. The following method sets the size of all
the light points in a pfGeoSet:
pfLPStateVal(lpstate,PFLPS_SIZE_ACTUAL, real-size);

real-size is a float that is multiplied times the distance. When real-size is less than 1.0, the
product of the distance and the real-size value is less than the distance. The revised
distance is used in calculating the apparent size of the light point; when the real-size is
less than 1.0, the light point appears larger than it otherwise would. The default real-size
value is 0.25.
Size Limitations

The apparent size of a light point can only range between the minimum and maximum
sizes specified by the following lines of code:
pfLPStateVal(lpstate,PFLPS_SIZE_MIN_PIXEL, min-size);
pfLPStateVal(lpstate,PFLPS_SIZE_MAX_PIXEL, max-size);

Limiting Size Calculations

To optimize rendering, you can avoid changing the size (glPointSize) of each light point
when the difference between the new and old sizes is less than a specified amount, called
the threshold. To set the threshold value, use the following method:
pfLPStateVal(lpstate, PFLPS_SIZE_DIFF_THRESH, threshold);

Enabling Perspective

To use real-size, you must first enable it using the following method:
pfLPStateMode(lpstate, PFLPS_SIZE_MODE, mode);

The mode value can be one of the following:
•

PFLPS_SIZE_MODE_ON makes real-size follow perspective so that light points
closer to the eye are rendered larger than those points farther away.

•

PFLPS_SIZE_MODE_OFF, the default value, disables perspective computations.

If perspective is not enabled, pfGeoSet (pfGSetPntSize) specifies the size of the
rendered light points.

578

007-1680-080

Callbacks

Fading
You can enhance the illusion of perspective by making the light point become more and
more transparent as it recedes from view using the following method:
pfLPStateVal(lpstate,PFLPS_TRANSP_PIXEL_SIZE, transp-size);

transp-size is a light point size. When the actual size is smaller than transp-size, the light
point becomes transparent.
Using fading to simulate perspective is often more realistic than shrinking the light point
size; fading avoids the aliasing problems that occur when light points become too small.
Fading Calculation

Fading is calculated as follows:
Max(clamp, 1-scale*(transp-size - computed-size)^exp)

You set the values in this argument in the following lines of code:
pfLPStateVal(lpstate, PFLPS_TRANSP_EXPONENT, exp);
pfLPStateVal(lpstate, PFLPS_TRANSP_SCALE, scale);
pfLPStateVal(lpstate, PFLPS_TRANSP_CLAMP, clamp);

Enabling Fading

To use fading, you must first enable it using the following method:
pfLPStateMode(lpstate, PFLPS_TRANSP_MODE, mode)

The mode value can be one of the following:
•

PFLPS_TRANSP_MODE_ON enables fading when the computed size is less than
transp-size.

•

PFLPS_TRANSP_MODE_OFF, the default, disables fading.

Callbacks
For each light point, two parameters are computed based on the location of the eye point,
the location of the light point, and the fog:

007-1680-080

579

18: Light Points

•

Alpha—Specifies the intensity and transparency of a light point. Non-transparent
light points have the maximum intensity given by their four color components.

•

Size—Specifies the diameter of each light point if the perspective mode,
PFLPS_SIZE_MODE, is enabled; otherwise, the size is constant for all of the light
points in a pfGeoSet.

Instead of accepting the calculations done by OpenGL Performer, you can use callback
functions to supply your own calculations. Callback functions can be completed at the
following times:
•

Before OpenGL Performer calculates the parameters, thus replacing the Performer
calculation completely.

•

After OpenGL Performer calculates the parameters, thus modifying Performer’s
result.

You enable, disable, and specify the kind of callback used in the following method:
pfLPStateMode(lpstate, PFLPS_CALLBACK_MODE, mode)

The mode value can be one of the following:
•

PFLPS_CALLBACK_MODE_OFF, the default, means that no callback is attached to
lpstate.

•

PFLPS_CALLBACK_MODE_PRE means that the callback is executed before
OpenGL Performer has calculated the parameters. Your callback function must
compute the size and alpha values for all the light points in the pfGeoSet.

•

PFLPS_CALLBACK_MODE_POST means that the callback is executed after
OpenGL Performer’s computation.

To install a callback on a pfLPointState, use the following method:
pfRasterFunc(lpstate, (void *) yourCallback(pfRasterData*),
void **userData);

userdata is a pointer to the following structure, which is given to your callback and the
function itself:
typedef struct {
pfLPointState
pfGeoSet
void
float

580

*lpstate;
*geoset;
*userData;
*sizes;

/*
/*
/*
/*

Read Only LPState */
Read Only GeoSet */
Provided when setting the callback */
Write Only - resulting sizes */

007-1680-080

Callbacks

float
} pfRasterData;

*alphas;

/* Write Only - resulting alphas */

geoset is the currently-preprocessed pfGeoSet. lpstate is the pfLPointState applied to that
pfGeoSet. userData is the same data provided when declaring the callback.
sizes and alphas are pre-allocated arrays that contain the callback function results used by
the DRAW process. If you use a pre-callback, you must provide a size and an alpha value
for every light point in the pfGeoSet. A negative alpha value indicates that the backcolor
must be used in place of the light point color.
Example 18-1 provides the skeleton of a raster callback. A more detailed example is on
the pfLPointState man page.
Example 18-1

Raster Callback Skeleton

void myCallback(pfRasterData *rasterData)
{
pfVec3* vertices;
unsigned short *vindex;
pfVec3* norms;
unsigned short *nindex;
int nbind;
pfFog *fog;
int fogEnabled = 0;
int i,n;
int sizeMode;
pfMatrix ViewMat, InvModelView;
/* get pointers to the geoset */
pfGetGSetAttrLists(rasterData->gset,PFGS_COORD3, &vertices, &vindex);
pfGetGSetAttrLists(rasterData->gset,PFGS_NORMAL3, &norms, &nindex);
nbind = pfGetGSetAttrBind(rasterData->gset,PFGS_NORMAL3);
/* get matrices */
pfGetViewMat(ViewMat);
pfGetInvModelMat(InvModelMat);
/* get the number of lights */
n = pfGetGSetNumPrims(rasterData->gset);
/* see if there’s fog */
fog = pfGetCurFog();
if (pfGetEnable(PFEN_FOG) && fog)
fogEnabled = 1;

007-1680-080

581

18: Light Points

/* get information on the lpstate */
sizeMode =
pfGetLPStateMode(rasterData->lpstate,PFLPS_SIZE_MODE);
........
........
/* do the computation */
for (i=0; ialphas[i] = ....
rasterData->sizes[i] = ....
}

Multisample, Size, and Alpha
On InfiniteReality, light points of a given size, up to 100 multisamples, have the same
number of multisamples even when the light points cross multiple pixels.
The intensity of a light point is defined by its alpha value. If pfTransparency is set to
PFTR_MS_ALPHA_MASK, the alpha value modifies the number of multisamples lit in
a light point. For example, if alpha= 0.5, size = 1.0, and the number of multisamples per
pixel is 8, the number of illuminated multisamples per the light point is 3.0, as shown in
Figure 18-4.

582

007-1680-080

Callbacks

Alpha = 0.5

Light point with
diameter = 1.0

8 multisamples
per pixel

Figure 18-4

Lit Multisamples

The area of a circle with diameter 1.0 is 0.785. The number of multisamples per pixel is
given in the frame buffer configuration. If there are 8 multisamples, only 0.785 of them,
about 6, are in the light point. Since alpha = 0.5, only half of the 6 multisamples are
actually lit.
pfCalligMultisample() tells pfCalligraphic how many multisamples are used in the
specified video channel. pfGetCalligMultisample() returns the current setting. Make
this call as soon as the information is known or changed.
Minimum Number of Multisamples

If your light points move, you must have at least two multisamples lit per light point; if
you only have one, the light point is either on or off. This condition creates flicker. With
two multisamples, the light point can transition to a different location with one
multisample on and the other off.

007-1680-080

583

18: Light Points

Reducing CPU Processing Using Textures
Light point computations are expensive. You can replace some of the computation by
using a table lookup mechanism and the texture hardware to get an alpha value for each
light point. This mechanism uses a precomputed texture and glTexGen to approximate:
•

fog attenuation

•

fading over distance

•

falloff attenuation when the eye point is not on the normal

Because only one texture per polygon is supported by the hardware, it is not possible to
approximate all three computations at once. The fading and the fog are combined in one
texture, however, so it is possible to select both at the same time.
The following methods specify the attenuations:
pfLPStateMode(lpstate, PFLPS_DIR_MODE, PFLPS_DIR_MODE_XXX)
pfLPStateMode(lpstate,PFLPS_TRANSP_MODE, PFLPS_TRANSP_MODE_XXX)
pfLPStateMode(lpstate, PFLPS_FOG_MODE, PFLPS_MODE_XXX)

XXX is one of the following:
•

TEX for texture lookup

•

ALPHA for CPU computation

Default values are equivalent to APLHA computation.
Tip: Falloff is the most expensive effect to compute on the CPU. For that reason, it is
better to use PFLPS_DIR_MODE_TEX if the point is not omnidirectional.
Only use textures for low quality light points; textures are approximations with
numerous limitations, including incorrect falloff attenuation if the emanation is not
symmetric and incompatibility with callbacks.

584

007-1680-080

Preprocessing Light Points

Preprocessing Light Points
To optimize your application, you should fork off a light point process to preprocesses
your light point computations. The light point process runs in parallel with the DRAW
process but on a different CPU. Preprocessing the light points does the following:
•

Computes the size and alpha of each light point and passes the result directly to the
DRAW process.

•

Executes callbacks attached to the pfLPointState.

To fork off a light point process, use the PFMP_FORK_LPOINT token in
pfMultiprocess() and call it before pfConfig().
Note: You can start a light point process in perfly using the -m option and adding 16
to your preferred multiprocess model.

Stage Configuration Callbacks
As with any other OpenGL Performer process, callback functions can configure the
process stages using the following:
pfStageConfigFunc(-1, PFPROC_LPOINT, ConfigLPoint);
pfConfigStage(-1, PFPROC_LPOINT | PFPROC_XXX ....);
pfChanTravFunc(chan, PFTRAV_LPOINT, LpointFunc);

pfStageConfigFunc() specifies a callback function, and pfConfigStage() triggers it at the
start of the current application frame; both are methods in pfConfig. Configuration
callbacks are typically used for process initialization; for example, they are used to do the
following:
•

Assign non-degrading priorities and lock processes to CPUs.

•

Download textures in the DRAW stage callback.

pfStageConfigFunc() identifies the OpenGL Performer stages, such as PFPROC_ISECT,
PFPROC_APP, and PFPROC_DBASE, to configure.

007-1680-080

585

18: Light Points

pfChanTravFunc

The pfChannel method, pfChanTravFunc(), sets a callback function for either the APP,
CULL, DRAW, or Light Point process using one of the following tokens: PFTRAV_APP,
PFTRAV_CULL, PFTRAV_DRAW, or PFTRAV_LPOINT, respectively.
User data that is passed to these functions by pfChanData() is allocated on a per-channel
basis by pfAllocChanData().

How the Light Point Process Works
This section explains how the light point process works. All the functions exist in the API;
however, all of the processing is done automatically by OpenGL Performer.
Note: Light point processing is not done automatically if your application is only using
the libpr process model.
If the light point process is enabled, a special bin, PFSORT_LPSTATE_BIN, is created to
sort all of the pfGeoSets that have a pfLPointState attached to their pfGeoState. This bin
is directly used as a display list, LPointBinDL, as shown in Example 18-2.
All bins, except PFSORT_LPSTATE_BIN, are handed to the DRAW process. The DRAW
process, after rendering everything in the other bins, renders a ring display list, called
DrawRingDL, as shown in Example 18-2. The ring display list contains a synchronization
mechanism: the DRAW process waits until it sees the PFDL_END_OF_FRAME token.
Example 18-2

Preprocessing a Display List - Light Point Process code

/* open the draw ring display list */
pfOpenDList(DrawRingDL);
/* preprocess the light points bin. */
/* so the results go in the DrawRingFL */
pfPreprocessDList(LPointBinDL,PFDL_PREPROCESS_LPSTATE);
/* Signal the end of the list to the Draw process */
pfAddDListCmd(PFDL_END_OF_FRAME);
/* close the draw display list */
pfCloseDList(DrawRingDL);

586

007-1680-080

Calligraphic Light Points

Calligraphic Light Points
Calligraphic light points are very bright lights that can be displayed only on
specially-equipped display systems. Displaying calligraphic light points requires the
following:
•

A calligraphic light point board (LPB) with a special device driver. The driver is not
part of the OpenGL Performer distribution.

•

A calligraphic display system.

•

Special cables running between the graphics pipe video synchronization and the
raster manager (RM) boards on the LPB.

•

A platform, such as an InifiniteReality, OnyxIR, or Onyx2/3, that has a visibility
(VISI) bus.

Note: If you are not running on a system that has a VISI bus or you do not have this
optional hardware, you are limited to raster light points, as supported by pfLPointState.
The functionality described in the remainder of this chapter is not available on your
system. Nevertheless, a program and database designed for calligraphic light points can
be simulated on a non-calligraphic system.
Unlike raster displays, calligraphic displays direct the display system’s electron beam at
specified places on the screen. By directing the beam at specified places for specified
durations, it is possible to produce extremely bright light sources.

Warning: It is possible to destroy your display system by allowing the electron
beam to remain too long on the same screen location either by allotting too long a time
or by an application hanging. Extreme caution is required.

007-1680-080

587

18: Light Points

Calligraphic Versus Raster Displays
Table 18-1 summarizes the differences between raster and calligraphic displays.
Table 18-1

Raster Versus Calligraphic Displays

Raster

Calligraphic

Requires no special hardware.

Requires special hardware, including a Light Point Board
(LPB), cables, and a calligraphic-enabled display system.
Applications using calligraphic light points must run on a
machine that has a VISI board.

The electron beam sweeps across and The beam lands only on those parts of the screen where
down the screen left-to-right and
calligraphic lights are located.
top-to-bottom.
The electron beam stays on each pixel The electron beams stays on pixels for a variable length of
the same amount of time.
time potentially producing exceedingly bright light
sources.
A black dot produces a black pixel.

A black dot produces nothing; it is invisible.

If more than one point is drawn at the Light points are added to whatever light is already falling
same location, only the last point
on the pixel. A calligraphic light does not hide another
drawn is visible.
calligraphic light.
Raster images are displayed within
set time intervals, for example, 60
times a second.

When raster and calligraphic are displayed, the
calligraphic light points are displayed in whatever time is
left after the raster image is scanned. For more
information, see “Display Modes” on page 589

No real dangers associated with
raster displays.

A hanging application, for example, can leave the electron
beam aimed at a single point on the screen and quickly
burn it out. The same result is true if you
programmatically light up a pixel for too long.

If the entire image is not drawn to the If all of the light points are not drawn, frames are not
buffer, frames are dropped until the dropped; some light points are just not drawn.
entire image is ready for display.
(If a raster image is repeated in successive frames, the light
points are also repeated.)

Whenever the calligraphic mode renders a black pixel to the screen, it is completely
transparent and the raster image shows through.

588

007-1680-080

Calligraphic Light Points

When a pixel on the screen is targeted for raster and calligraphic light, both the raster and
calligraphic light points are displayed at the same pixel thus making it large (raster) and
bright (calligraphic).
Display Modes

A calligraphic display system can run in three modes:
•

calligraphic-only

•

mixed mode

•

raster-only

In calligraphic-only mode, only calligraphic points can be rendered by the display
system.
In mixed mode, both raster and calligraphic images are rendered on the same display
system; the raster image is displayed first and the calligraphic image is displayed in
whatever time remains before the vertical sync. This mode requires a special video
format used by the Video Format Compiler available on the SGI web site at
http://www.sgi.com/software/vfc/.
You can combine a calligraphic-only display with a raster-only display on the same video
channel; the effect is to give the full frame to the calligraphic display so that you can
render the maximum number of light points. It is also more expensive and sometimes not
convenient for mechanical reasons.
Light point modes are specified by pfLPStateMode() and one of the following tokens:
•

PFLPS_DRAW_MODE_RASTER, the default mode, forces the light points to be
raster even if the system has a calligraphic display.

•

PFLPS_DRAW_MODE_CALLIGRAPHIC enables the rendering of calligraphic
light points on calligraphic display systems.

Maximum Number of Calligraphic Lights

The maximum number of calligraphic lights that can be displayed is related to the
following:

007-1680-080

•

raster display time

•

duration of the calligraphic display time

•

time spent jumping from one calligraphic light point to another

589

18: Light Points

The maximum number of calligraphic lights that can be displayed is inversely
proportional to the raster display time. For example, in a mixed mode, if the screen
refreshes every 1/60th of a second, the calligraphic display time is 1/60th of a second
minus the time it takes for the raster mode to draw its image on the screen. The shorter
the raster display time, the more calligraphic lights that can be displayed.
At night, it is possible to reduce the time for the raster display, which has the effect of
reducing the global raster brightness. Reducing the raster display time increases the
maximum number of calligraphic light points that can be displayed. To reduce the raster
display time, you need to do the following:
•

Two different video formats made using the Video Format Compiler.

•

Preprogram the projector so it recognizes the format change when it happens on the
fly by the application.

Changing the video format is done using the XSGIvc extension.
The maximum number of calligraphic lights that can be displayed is inversely
proportional to the draw time of the calligraphic light points. Unlike raster lights, you
can control the length of time the electron beam hits a specified location on the display
system; the longer the duration, the brighter the light. Also, the longer the duration, the
less time is left for drawing other calligraphic light points.
Finally, the time it takes the electron beam to jump between calligraphic light points also
adversely affects the maximum number of light points that can be displayed; the more
time spent jumping between calligraphic light points, the fewer light points that can be
displayed.

LPB Hardware Configuration
A Light Point Board (LPB) is a circuitry card that enables the rendering of calligraphic
lights by providing the interface with a calligraphic display. The LPB is configured as
follows:

590

•

Connected to one graphics pipeline only.

•

Connected to all the video channels produced by a single pipe.

•

Connected to all (1,2 or 4) raster manager (RM) boards of its pipe.

007-1680-080

Calligraphic Light Points

Color
Focus
Exposure time
Quality and position
of light point

Light Point
process

CPU

VME bus

CPU

Sync signals,
including
Swap Ready

Z-buffer
comparison

Draw
process

VISI bus

Channel 0...7
LPB
Graphics
pipeline
Geometry
Engine

Calligraphic
light points

Raster
Manager
Video
Channel 0...7

RGB
Channel 0

Figure 18-5

Fiber
optic
cable
Vid

sig

nal

Calligraphic Hardware Configuration

The configuration in Figure 18-5 shows the following:

007-1680-080

•

A CPU is used for the light point process so the computation and the
communication to the LPB through the VME bus are done in parallel to the DRAW
process. It is mandatory to start a light point process for calligraphic; see
“Preprocessing Light Points” on page 585.

•

The VME bus transfers to the LPB all of the calligraphic light point information,
including the color, focus, exposure time, quality, and position of each calligraphic
light point. The only information not transferred by the VME bus is occlusion
information, which is supplied by the VISI bus.

•

The VISI (visibility) bus specifies whether all, part, or none of a calligraphic light
point is displayed. A calligraphic light might be partially displayed, or not at all, if a
geometry in the scene is between the light point and the viewer. Each light point has

591

18: Light Points

a unique ID; this ID is used to match the information in the VISI bus with the correct
light point.
The VISI bus is a connector on each RM board. The visibility information is
available only to the LPB.
Note: The LPB board may be used in a system without a VISI bus (systems prior to
InfiniteReality), in which case no Z-buffer information is given to the board; so, all
light points are 100% visible.
•

The LPB uses the vertical and horizontal (not the composite) synchronization
signals to trigger the calligraphics display. Use ircombine to set the Hsync in place
of the composite sync.

•

The LPB receives the Swap Ready signal when the raster display has completed
drawing to the display buffer; so, it should also swap its internal buffers.
If the LPB does not get a Swap Ready signal, the LPB redisplays the same
calligraphic light points; since the raster frame is repeated, the calligraphic lights
must remain unchanged. Do not forget to connect the SwapReady signal to the light
point board even if you are using a single pipe configuration.

•

The LPB receives the VISI and VME bus light point information and combines it
and send the result to the calligraphic display.

Visibility Information
All the calligraphic light point information is computed by the light point process and
goes directly to the LPB. The only missing information is knowing how much of each
light point can be seen. To compute that value, the following events take place:
1.

A footprint of the calligraphic light point is sent to the graphic pipe.

2. The footprint is compared against the Z-buffer.
3. The result of the test is sent to the LPB using the VISI bus.
The graphic pipe takes great care that the number of multisamples covered is a constant
wherever the light point is. The number of multisamples is not constant after the
footprint covers more than 100 multisamples, however. A footprint can cover numerous
pixels but is limited to 256 multisamples by the LPB.

592

007-1680-080

Calligraphic Light Points

pfCalligZFootPrintSize() sets the diameter of the footprint and
pfGetCalligZFootPrintSize() returns the diameter. The number of multisamples
covered by a footprint is equal to the following:
(n x size2)/(4 x ms)

ms is the number of multisamples per pixel.

Required Steps For Using Calligraphic Lights
To use calligraphic light points, you have to configure the channels on the LPB board and
the corresponding channels using ircombine. ircombine operates on the following:
•

video format combinations

•

descriptions of raster sizes and timings used on video outputs

•

configuration of the underlying frame buffer

You must also enable the channels on the LPB board before starting the application.
OpenGL Performer does not provide direct access to the LPB drivers. You can, however,
write a program that does.
Now, in OpenGL Performer a few steps are still necessary.
1.

Check and open the LPB on each pipe, as follows:
pfQueryFeature(PFQFTR_CALLIGRAPHIC, &q)
pfCalligInitBoard(pipe)

2. Start a light point process to process the light point calculations, as follows:
pfMultiprocess(PFMP_APP_CULL_DRAW |
pfConfig();

PFMP_FORK_LPOINT);

3. Initialize the OpenGL Performer stages, as follows:
pfStageConfigFunc(-1,PPROC_LPOINT,ConfigLPoint)
pfFrame();

4. Set the callback function in the light point process, as follows:
pfChanTravFunc(chan, PFTRAV_LPOINT, LpointFunc);

5. Enable the calligraphic display on all of the channels, as follows:
pfChanCalligEnable(chan[i], 1);

007-1680-080

593

18: Light Points

6. Synchronize the VME and VISI bus signals on the LPB, as follows:
pfCalligSwapVME(pipe);
pfCalligSwapVME(pipe);

You now have calligraphic light points in your application if you have calligraphic light
points in your database.
Customizing LPB Initialization

Instead of using the default initialization, OpenGL Performer provides a set of functions
that allow your application to customize the initialization of the LPB.
pfInitBoard() opens the LPB device and retrieves the current configuration. This
function returns TRUE if successful, FALSE otherwise. You can also use pfIsBoardInit()
to determine if the board has been initialized. Each LPB must be initialized before calling
pfConfig().
pfCalligCloseBoard() closes the device.
Note: The board number and the pipe number are the same because there is only one
board per pipe.
To access to the LPB directly, return its ID using pfGetCalligDeviceID(). The ID allows
you to make direct calls to the LPB driver.
pfGetCalligInfo() returns a pointer to the configuration information structure
maintained by the LPB driver. To use this structure, LPB_info, you must include the
driver lpb.h file before any OpenGL Performer include files.
Note: lpb.h is not distributed with OpenGL Performer; it is part of the LPB driver
distribution.
Once a board is initialized, you can find out how much memory is available and allocate
it among all of the enabled channels on the pipe. You can return the total amount of LPB
memory in bytes by calling pfGetCalligBoardMemSize().

594

007-1680-080

Calligraphic Light Points

By default, each channel receives the same amount of memory. To set up a different
partitioning of memory, use pfCalligPartition().
You can attach a pfCalligraphic to a specific pfChannel using the following code:
pfCalligraphic *callig = pfNewCallig(arena)
pfCalligChannel(callig, pipe, chan)

callig can then be attached to the pfPipeVideoChannel using pfPVChanCallig().
You can also set the pfCalligraphic on individual pfChannels using pfChanCallig(). You
must enable calligraphic light point processing on the specified channels so the
GangSwap mechanism is correctly handled by OpenGL Performer, as follows:
pfChanCalligEnable(chan, 1);

Note: The video channel number does not have to be the same channel number as the
calligraphic board.
pfCalligWin() changes the resolution of the X and Y data accepted by the projector. The
default values are for an EIS projector (2^16). If you have a conversion interface between
the LPB and your projector, the scaling may already be done by the conversion interface.
You can use pfCalligWin() to display calligraphic points on only part of the screen or to
make multiple viewports.
pfCalligXYSwap() switches the X and Y axes in the display. You can also reverse the X
or Y axes by giving negative width and height values in pfCalligWin().

Accounting for Projector Differences
Some display systems, such as the EIS projector, can calculate the following:
•

Slew values, the time for the electron beam to go from one calligraphic point to
another.

•

Gamma correction, which takes into account that projectors differ in the colors they
project.

If your system cannot perform those calculations, the light point board can calculate
those values for your application.

007-1680-080

595

18: Light Points

Slew Values

A slew table is two dimensional; it gives the time in nanoseconds it takes the electron
beam to go from one calligraphic point to another on one axis. A default, generic slew
table, which contains conservative values, is loaded in the LPB when it is initialized.
The longer the slew time, the longer the electron beam has to reach a calligraphic light
point on the display; any consequent wobble can dampen out in that time providing a
very stable light point. Conversely, longer slew times subtract from the total time in
which calligraphic light points can be drawn. Consequently, longer slew times could
mean that fewer calligraphic points are drawn per frame.
There are eight slew tables defined by the pfCalligSlewTableEnum. For each axis, there
are three slew tables: one table for high-quality (very stable) light points, one table for
medium-quality (slightly shorter slew times) light points, and another for low-quality
(very short slew times) light points.
You specify the drawing quality using pfLPStateVal() with one of the following mode
values:
•

PFLPS_QUALITY_MODE_HIGH

•

PFLPS_QUALITY_MODE_MEDIUM

•

PFLPS_QUALITY_MODE_LOW

Two other tables are used when the defocus value changes in between two points; one
table is used for high-quality light points, another table for medium- and low-quality
light points.
typedef enum {
pfXSlewQuality0 = 0, /* High quality on the X axis */
pfXSlewQuality1 = 1, /* Medium quality on the X axis */
pfXSlewQuality2 = 2, /* Low quality on the X axis */
pfYSlewQuality0 = 3, /* High quality on the Y axis */
pfYSlewQuality1 = 4, /* Medium quality on the Y axis */
pfYSlewQuality2 = 5, /* Low quality on the Y axis */
pfDefocusQuality0 = 6,/* High quality if focus change */
pfDefocusQuality1 = 7 /* Medium and Low quality if focus change */
} pfCalligSlewTableEnum;

To load or upload customized slew tables, use the following methods:

596

•

pfCalligDownLoadSlewTable() downloads a specified slew table into the LPB.

•

pfCalligUpLoadSlewTable() returns a slew table.

007-1680-080

Calligraphic Light Points

Color Correction

Each projector has its own color characteristics; a light point that appears blue on one
projector might appear aqua on another. To color correct the projected images, the light
point board maintains one gamma table per channel. Each gamma table consists of three
one-dimensional tables, one for each color component.
typedef enum {
pfRedGammaTable = 0,
pfGreenGammaTable = 1,
pfBlueGammaTable = 2
} pfCalligGammaTableEnum;

The default value provided by OpenGL Performer is a linear ramp that can be modified
with the following methods:
•

pfCalligDownLoadGammaTable() downloads a specified gamma table into the
LPB.

•

pfCalligUpLoadGammaTable() returns a gamma table.

Callbacks
Like raster lights, calligraphic light points can have a callback function attached to the
pfLPointState that can occur before (PRE) or after (POST) the light point processing. The
calligraphic and raster callback functions compute different parameters.
If the stress test determines that a calligraphic light point is not going to be drawn as a
raster light point, the calligraphic callback is not be called; the raster callback is called
instead if set. For more information about the stress test, see “Significance” on page 600.
To install a calligraphic callback on a pfLPointState, you use a line of code similar to the
following:
pfCalligFunc(lpstate, (void *) yourCallback(pfCalligData*),
void **userData);

userData is a user pointer given to your callback identifying the callback function and the
following structure:
typedef struct {
pfLPointState
pfGeoSet
void

007-1680-080

*lpstate; /* Read Only LPState */
*geoset;
/* Read Only GeoSet */
*userData; /* Provided when setting the callback */

597

18: Light Points

unsigned short
int
pfVec3
float
float
float
pfCalligData;
} pfRasterData;

*index;
/*
*n;
/*
*coords2D; /*
*intensity;/*
**focus;
/*
**drawTime;/*

Read Write
Read Write
Read Write
Write Only
Write Only
Write Only

index visible lpoints */
# of visible lpoints */
screen space X,Y,Z */
resulting intensity */
optional (de)focus */
optional drawTime */}

geoset is the pfGeoSet that is currently preprocessed. lpstate is the pfLPointState applied
to that pfGeoSet. userData is the same as the data provided when declaring the callback.
index is a preallocated vector that points to light points that are visible. Even in a PRE
callback the light points are projected on the screen before the user callback is called; the
points outside of the screen are not in the index vector. See the pfLPointState man
page on how to use the index vector in a callback.
n is a pointer to the number of elements in the index vector.
coords2D contains the coordinates of the light points on the screen, including the
Z coordinates in screen space. The original coordinates can be accessed through the
pfGeoSet. It is valid to change the values in coord2D in the callback, for example, to align
the points on a grid.
intensity is a pre-allocated vector that contains the intensity of individual light points. A
PRE callback must compute intensity.
focus is a NULL pointer. It is possible to provide an individual focus value for each point
by doing the following:
1.

Allocating from the arena an array of floats of size n.

2. Filling the array with the focus values.
3. Setting focus to point to that array.
You should not allocate an array in real-time; instead, allocate temporary memory
beforehand and use userData to pass the memory address to the callback.
drawTime is a NULL pointer. With it you can provide an array of floats that give a draw
time for each light point.

598

007-1680-080

Calligraphic Light Points

Frame to Frame Control
Before calling pfLPoint() in the light point process callback function, you can change the
parameters in a pfCalligraphic object on a frame-by-frame basis, as shown in
Example 18-3.
Example 18-3

Setting pfCalligraphic Parameters

myLPointFunc(pfChannel *chan, void *data)
{
pfCalligraphic *calligraphic = pfGetCurCallig();
if (calligraphic != NULL)
{
pfCalligFilterSize(calligraphic,FilterSizeX,FilterSizeY);
pfCalligDefocus(calligraphic,Defocus);
pfCalligRasterDefocus(calligraphic,rasterDefocus);
pfCalligDrawTime(calligraphic, DrawTime);
pfCalligStress(calligraphic, Stress)
}
pfLPoint();
}

FillterSizeX and FilterSizeY set the debunching distances along each axis. A filter size of 0
disables the debunching along that axis. Debunching can also be disabled on the
pfLPointState. For more information about debunching, see “Debunching” on page 600.
Defocus sets the defocus applied to all calligraphic light points, unless a callback function
returns a defocus array. In that case, each light point has a callback defocus value
regardless of the one set on the pfCalligraphic object.
In a raster-plus-calligraphic video mode, a calligraphic system usually has the capability
to apply a global defocus to the raster image, which is specified by the rasterDefocus
parameter passed to the projector. This parameter affects the entire image, not just the
raster light points.
For more information about defocus, see “Defocussing Calligraphic Objects” on
page 601.
DrawTime is the time, in nano seconds, in which all calligraphic light points must be
drawn, unless a callback returns a DrawTime array, which controls individual draw
times. The overall effect of changing the draw time is to reduce or increase the intensity
of all calligraphic light points.

007-1680-080

599

18: Light Points

The Stress value is compared against the significance value of a pfLPointState to determine
whether the light points are displayed as calligraphic or raster. The Stress value should
be as stable as possible to avoid having points going from raster to calligraphic or from
calligraphic to raster each frame. For more information about stress, see “Significance”
on page 600.

Significance
If there is not enough time to draw all of the calligraphic light points, some calligraphic
light points can either not be drawn or, instead, drawn as raster light points, depending
on the stress value of the calligraphic light point, described in “Frame to Frame Control”
on page 599, and the significance value of the pfLPointState.
The stress value of a calligraphic light point is compared against the significance value of
the pfLPointState. If the significance is greater or equal than the stress, the light is
rendered as a calligraphic; otherwise, the light is rendered as a raster light.
To set the significance, use pfLPStateVal(lpstate, PFLPS_SIGNIFICANCE, significance).

Debunching
Debunching eliminates some calligraphic points when they occur close together on the
display system. Bunched calligraphic light points can create a heap effect, or worse, burn
the display system where there is a group of calligraphic light points.
The debunching distances are set in pfCalligraphic on the X and Y axis, as described in
“Frame to Frame Control” on page 599.
If two points in a pfGeoSet are within the debunching distance, the point with the lowest
intensity is not rendered.
If you do not want a pfLPointState to be affected by debunching, you can disable it using
pfLPStateMode() with the PFLPS_DEBUNCHING_MODE_OFF mode;
PFLPS_DEBUNCHING_MODE_ON, the default, enables debunching.

600

007-1680-080

Using pfCalligraphic Without pfChannel

Defocussing Calligraphic Objects
Defocussing a calligraphic light point is often done to produce a specific lighting effect,
for example, to simulate rainy conditions, light points should appear defocussed. The
defocus is set within pfCalligraphic (see “Frame to Frame Control” on page 599) for all
light points for one Frame, but it is possible to clamp the defocussing values by setting a
min_defocus and a max_defocus value in the pfLPState using the following methods:
pfLPStateVal(lpstate,PFLPS_MIN_DEFOCUS, min_defocus);
pfLPStateVal(lpstate, PFLPS_MAX_DEFOCUS, max_defocus);

Using pfCalligraphic Without pfChannel
pfCalligraphic is a libpr object; so, programs based only on libpr can have access to
pfCalligraphic without having access to a pfChannel, which is a libpf object. This
section describes what pfChannel does automatically with pfCalligraphic objects.
The LPB is connected to the VME bus and the VISI bus. Both buses contain information
for calligraphic light points buffered in the LPB. VISI and VME busses load their
information into the LPB buffer. The SwapReady connection to the graphic pipe
synchronizes the activity on the VISI bus; SwapReady tells the board when
glXSwapBuffers() is called, and that at the next VSync, the next frame should be
displayed.
pfCalligSwapVME() performs the same synchronization for the VME bus; whenever
you call pfSwapPWinBuffers(), you must first call pfCalligraphicSwapVME();
otherwise, the light point board will not be synchronized with the rest of the system.
Tip: To resynchronize the LPB, make two consecutive calls to pfCalligSwapVME(); for
example, perfly makes these calls after the OpenGL Performer logo has been
displayed.

Note: These synchronization mechanisms are handled automatically in libpf, unless
you override the channel swap buffer and call pfSwapPWinBuffers().

007-1680-080

601

18: Light Points

Timing Issues
The SwapReady signal should always occur after a VME Swap signal. If this does not
happen, the light point board starts a TooLate timer and two things may happen:
•

Assuming the VME Swap signal has been lost, the LPB starts to draw the buffer .
In this instance, not all of the calligraphic lights are rendered. pfCalligraphic
handles this exception by making sure the LPB buffer always contains an
end-of-buffer token at the end of the valid data.

•

The VME Swap signal is received shortly after the Swap Ready, making the LPB
behave normally.

The LPB needs some time before it can accept new information from the VME, and the
VISI bus needs some time after it has received the corresponding swap command.
pfWaitForVmeBus() and pfWaitForVisiBus() allow the application to wait for the board
to get ready before sending it new information.

Light Point Process and Calligraphic
The light point process is the same whether the light points are raster or calligraphic. The
only difference is that, with calligraphic light points, pfCalligraphic has to be selected
before calling pfPreprocessDList, which is done automatically by pfChannel.
pfSelectCalligraphic() selects the channel and LPB board to which the calligraphic light
points are sent. pfGetCurCalligraphic() returns the current value set by
pfSelectCalligraphic().

Debugging Calligraphic Lights on Non-Calligraphic Systems
If you are developing on a non-calligraphics-enabled system but would like to see the
effects of your pfCalligraphics programming, you can set the environment variable
PF_LPOINT_BOARD. In this mode, the LPB is simulated, the calligraphic computations
are performed, and raster light points are displayed in place of the calligraphic lights
with the following limitations:

602

•

Calligraphic defocus has no effect.

•

A calligraphic light point size is defined with the Z-footprint.

007-1680-080

Calligraphic Light Example

Calligraphic Light Example
Example 18-4 shows a sample implementation of calligraphic lights. You can find the
source code in perf/sample/pguide/libpf/C/callig.c.
Example 18-4
#include
#include
#include
#include

Calligraphic Lights

static NumScreens=1;
static NumPipes=1;
static
static
static
static

void
void
void
void

ConfigPipeDraw(int pipe, uint stage);
OpenPipeWin(pfPipeWindow *pw);
OpenXWin(pfPipeWindow *pw);
DrawChannel(pfChannel *chan, void *data);

/*
* Usage() -- print usage advice and exit. This
*
procedure is executed in the application process.
*/
static void
Usage (void)
{
pfNotify(PFNFY_FATAL, PFNFY_USAGE, “Usage: multipipe file.ext
...\n”);
exit(1);
}
int
main (int argc, char *argv[])
{
float
t = 0.0f;
pfScene
*scene;
pfPipe
*pipe[4];
pfChannel
*chan[4];
pfNode root;
pfSphere sphere;
int loop;

007-1680-080

603

18: Light Points

/* Initialize Performer */
pfInit();
pfuInitUtil();
/* specify the number of pfPipes */
/* Configure and open GL windows */
if ((NumScreens = ScreenCount(pfGetCurWSConnection())) > 1)
{
NumPipes = NumScreens;
}
pfMultipipe (NumPipes);
/* Initialize Calligraphic HW */
for (loop=0; loop < NumPipes; loop++)
{
int q;
pfQueryFeature(PFQFTR_CALLIGRAPHIC, &q);
if (!q)
{
pfNotify(PFNFY_NOTICE,PFNFY_RESOURCE, “Calligraphic points
are NOT supported on this platform”);
}
if (pfCalligInitBoard(loop) == TRUE)
{
/* get the memory size */
pfNotify(PFNFY_NOTICE, PFNFY_RESOURCE,“StartCalligraphic:
board(%d) has %d Bytes of memory”,0, pfGetCalligBoardMemSize(loop));
}
else
{
pfNotify(PFNFY_NOTICE, PFNFY_RESOURCE, “Could not init
calligraphic board %d”, loop);
}
}
/* Force Multiprocessor mode to use the light point process */
pfMultiprocess(PFMP_APP_CULL_DRAW |

PFMP_FORK_LPOINT);

/* Load all loader DSO’s before pfConfig() forks */
if (argc > 1)
pfdInitConverter(argv[1]);

604

007-1680-080

Calligraphic Light Example

/* Configure multiprocessing mode and start parallel processes.*/
pfConfig();
/* Append to PFPATH additional standard directories where geometry
and textures exist */
scene = pfNewScene();
if (argc > 1)
{
if (!(getenv(“PFPATH”)))
pfFilePathv(“.", "./data", "../data", "../../data",
"/usr/share/Performer/data”, NULL);
/* Read a single file, of any known type. */
if ((root = pfdLoadFile(argv[1])) == NULL)
{
pfExit();
exit(-1);
}
/* Attach loaded file to a pfScene. */
pfAddChild(scene, root);
/* determine extent of scene’s geometry */
pfGetNodeBSphere (scene, &bsphere);
/* Create a pfLightSource and attach it to scene. */
pfAddChild(scene, pfNewLSource());
}

for (loop=0; loop < NumPipes; loop++)
pfPipeWindow *pw;
char str[PF_MAXSTRING];
pipe[loop] = pfGetPipe(loop);
pfPipeScreen(pipe[loop], loop);
pw = pfNewPWin(pipe[loop]);
pfPWinType(pw, PFPWIN_TYPE_X);
sprintf(str, “OpenGL Performer - Pipe %d”, loop);
pfPWinName(pw, str);
if (NumScreens > 1)
{
pfPWinOriginSize(pw, 0, 0, 300, 300);
} else

007-1680-080

605

18: Light Points

pfPWinOriginSize(pw, (loop&0x1)*315, ((loop&0x2)>>1)*340, 300,
300);
pfPWinConfigFunc(pw, OpenPipeWin);
pfConfigPWin(pw);
}

/* set up optional DRAW pipe stage config callback */
pfStageConfigFunc(-1 /* selects all pipes */, PFPROC_DRAW /* stage
bitmask */, ConfigPipeDraw);
pfConfigStage(-1, PFPROC_DRAW);
pfFrame();
/* Create and configure pfChannels - by default, channels are
placed in the first window on their pipe */
for (loop=0; loop < NumPipes; loop++)
{
chan[loop] = pfNewChan(pipe[loop]);
pfChanScene(chan[loop], scene);
pfChanFOV(chan[loop], 45.0f, 0.0f);
pfChanTravFunc(chan[loop], PFTRAV_DRAW, DrawChannel);
pfChanCalligEnable(chan[loop], 1);
/* synchronize the lpoint board with the swap ready signal */
pfCalligSwapVME(loop);
pfCalligSwapVME(loop);
}
/* Simulate for twenty seconds. */
while (t < 30.0f)
{
pfCoord view;
float
s, c;
/* Go to sleep until next frame time. */
pfSync();
/* Initiate cull/draw for this frame. */
pfFrame();
pfSinCos(45.0f*t, &s, &c);
pfSetVec3(view.hpr, 45.0f*t, -10.0f, 0);

606

007-1680-080

Calligraphic Light Example

pfSetVec3(view.xyz, 2.0f * bsphere.radius * s, -2.0f *
bsphere.radius *c, 0.5f * bsphere.radius);
for (loop=0; loop < NumPipes; loop++)
pfChanView(chan[loop], view.xyz, view.hpr);
}
/* Terminate parallel processes and exit. */
pfExit();
return 0;
}
/* ConfigPipeDraw() -* 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.
*/
static void
ConfigPipeDraw(int pipe, uint stage)
{
pfPipe *p = pfGetPipe(pipe);
int x, y;
pfNotify(PFNFY_NOTICE, PFNFY_PRINT, “Initializing stage 0x%x of
pipe %d on screen %d, connection \”%s\””, stage,
pipe,pfGetPipeScreen(p),pfGetPipeWSConnectionName(p));
pfGetPipeSize(p, &x, &y);
pfNotify(PFNFY_NOTICE, PFNFY_PRINT, “Pipe %d size: %dx%d”, pipe,
x,y);
}
/*
* OpenPipeWin() -- create a GL window: set up the
*
window system and OpenGL Performer. This
*
procedure is executed for each window in the draw process
*
for that pfPipe.
*/

void
OpenXWin(pfPipeWindow *pw)
{

007-1680-080

607

18: Light Points

/* -1 -> use default screen or that specified by shell DISPLAY
variable */
pfuGLXWindow
*win;
pfPipe *p;
p = pfGetPWinPipe(pw);
if (!(win = pfuGLXWinopen(p, pw, “TESTIN”)))
pfNotify(PFNFY_FATAL, PFNFY_RESOURCE,“OpenXPipeline: Could not
create GLX Window.”);
win = win; /* suppress compiler warn */

}
static void
OpenPipeWin(pfPipeWindow *pw)
{
pfPipe *p;
pfLight *Sun;
p = pfGetPWinPipe(pw);
/* open the window on the specified screen. By default,
* if a screen has not yet been set and we are in multipipe mode,
* a window of pfPipeID i will now open on screen i
*/
pfOpenPWin(pw);
pfNotify(PFNFY_NOTICE, PFNFY_PRINT,
PipeWin of Pipe %d opened on screen %d”,
pfGetId(p),pfGetPipeScreen(p));
/* create a light source in the “south-west” (QIII) */
Sun = pfNewLight(NULL);
pfLightPos(Sun, -0.3f, -0.3f, 1.0f, 0.0f);
}
static void
DrawChannel (pfChannel *channel, void *data)
{
static int first = 1;
if (first)
{
first = 0;

608

007-1680-080

Calligraphic Light Example

pfNotify(PFNFY_NOTICE,PFNFY_PRINT,”Calligraphics: 0x%p”,
pfGetChanCurCallig(channel));
}
/* erase framebuffer and draw Earth-Sky model */
pfClearChan(channel);
/* invoke Performer draw-processing for this frame */
pfDraw();
}

007-1680-080

609

Chapter 19

19. Math Routines

This chapter describes the OpenGL Performer math routines. Math routines let you
create, modify, and manipulate vectors, matrices, line segments, planes, and bounding
volumes such as spheres, boxes, and cylinders.

Vector Operations
A basic set of mathematical operations is provided for setting and manipulating floating
point vectors of length 2, 3, and 4. The types of these vectors are pfVec2, pfVec3, and
pfVec4, respectively. The components of a vector are denoted by PF_X, PF_Y, PF_Z, and
PF_W with indices of 0, 1, 2, and 3, respectively. In the case of 4-vectors, the PF_W
component acts as the homogeneous coordinate in transformations.
OpenGL Performer supplies macro equivalents for many of the routines described in this
section. Inlining the macros instead of calling the routines can substantially improve
performance. The C++ interface provides the same, fast performance as the inlined
macros.
Table 19-1 lists the routines, what they do (in mathematical notation), and the macro
equivalents (where available) for working with 3-vectors. Most of the same operations
are also available for 2-vectors and 4-vectors, substituting “2” or “4” for “3” in the routine
names. The only operations not available for 2-vectors are vector cross-products, point
transforms, and vector transforms; the only operations unavailable for 4-vectors are
vector cross-products and point transforms, that is, there are no such routines as
pfCrossVec2(), pfCrossVec4(), pfXformPt2(), pfXformPt4(), or pfXformVec2().

007-1680-080

611

19: Math Routines

Note: For the duration of this chapter, we use the following convention for denoting
one-letter variables: bold lowercase letters represent vectors and bold uppercase letters
represent matrices. “x” indicates cross product, “.” denotes dot product, and vertical bars
indicate the magnitude of a vector.]
Table 19-1

612

Routines for 3-Vectors

Routine

Effect

Macro Equivalent

pfSetVec3(d, x, y, z)

d = (x, y, z)

PFSET_VEC3

pfCopyVec3(d, v)

d=v

PFCOPY_VEC3

pfNegateVec3(d,v)

d = -v

PFNEGATE_VEC3

pfAddVec3(d, v1, v2)

d = v1 + v2

PFADD_VEC3

pfSubVec3(d, v1, v2)

d = v1 - v2

PFSUB_VEC3

pfScaleVec3(d, s, v)

d = sv

PFSCALE_VEC3

pfAddScaledVec3(d, v1, s, v2)

d = v1 + sv2

PFADD_SCALED_VEC3

pfCombineVec3(d,s1,v1,s2,v2)

d = s1v1 + s2v2

PFCOMBINE_VEC3

pfNormalizeVec3(d)

d = d/|d|

none

pfCrossVec3(d, v1, v2)

d = v1 x v2

none

pfXformPt3(d, v, m)

d = vM, where v = (vx, vy, vz,) and M none
is the 4x3 submatrix.

pfFullXformPt3(d, v, M)

d = vM/dw, where v = (vx, vy, vz, 1)

none

pfXformVec3(d, v, M)

d = vM, where v = (vx, vy, vz, 0)

none

pfDotVec3(v1, v2)

v1 . v2

PFDOT_VEC3

pfLengthVec3(v)

|v|

PFLENGTH_VEC3

pfSqrDistancePt3(v1, v2)

|v2 - v1|2

PFSQR_DISTANCE_PT3

pfDistancePt3(v1, v2)

|v2 - v1|

PFDISTANCE_PT3

007-1680-080

Matrix Operations

Table 19-1

Routines for 3-Vectors (continued)

Routine

Effect

Macro Equivalent

pfEqualVec3(v1, v2)

Returns TRUE if v1 = v2 and FALSE, PFEQUAL_VEC3
otherwise.

pfAlmostEqualVec3(v1,v2,tol)

Returns TRUE if each element of v1 PFALMOST_EQUAL_
is within tol of the corresponding
VEC3
element of v2 and FALSE, otherwise.

Matrix Operations
A pfMatrix is a 4x4 array of floating-point numbers that is used primarily to specify a
transformation in homogeneous coordinates (x, y, z, w). Transforming a vector by a
matrix means multiplying the matrix on the right by the row vector on the left.
Table 19-2 describes the OpenGL Performer mathematical operations that act on
matrices.
Table 19-2

Routines for 4x4 Matrices

Routine

Effect

Macro Equivalent

pfMakeIdentMat(d)

D = I.

PFMAKE_IDENT_MAT

pfMakeVecRotVecMat(d,v1,v2)

D = M such that v2 = v1M.

none

v1, v2 are normalized.

007-1680-080

pfMakeQuatMat(d, q))

D = M, where M is the rotation of none
the quaternion q.

pfMakeRotMat(d, deg, x, y, z)

D = M, where M rotates by deg
around (x, y, z).

pfMakeEulerMat(d, h, p, r)

D = RPH, where R, P, and H are none
the transforms for roll, pitch, and
heading.

pfMakeTransMat(d, x, y, z)

D = M, where M translates by
(x, y, z).

pfMakeScaleMat(d, x, y, z)

D = M, where M scales by (x, y, z). PFMAKE_SCALE_MAT

none

PFMAKE_TRANS_MAT

613

19: Math Routines

Table 19-2

614

Routines for 4x4 Matrices (continued)

Routine

Effect

Macro Equivalent

pfMakeCoordMat(d, c)

D = M, where M rotates by (h, p, none
r) and translates by (x, y, z) with h,
p, r, x, y, and z all specified by c.

pfGetOrthoMatQuat(s, q)

Returns in q a quaternion with
the rotation specified by s.

none

pfGetOrthoMatCoord(s, d)

Returns in d the rotation and
translation specified by s.

none

pfSetMatRow(d, r, x, y, z, w)

Set rth row of D equal to (x, y, z,
w).

PFSET_MAT_ROW

pfGetMatRow(m, r, x, y, z, w)

(*x, *y, *z, *w) = rth row of M.

PFGET_MAT_ROW

pfSetMatCol(d, c, x, y, z, w)

Set cth column of D equal to (x, y, PFSET_MAT_COL
z, w).

pfGetMatCol(m, c, x, y, z, w)

(*x, *y, *z, *w) = cth column of M. PFGET_MAT_COL

pfSetMatRowVec3(d, r, v)

Set rth row of D equal to v.

PFSET_MAT_ROWVEC3

pfGetMatRowVec3(m, r, d)

d = rth row of M.

PFGET_MAT_ROWVEC3

pfSetMatColVec3(d, c, v)

Set cth column of D equal to v.

PFSET_MAT_COLVEC3

pfGetMatColVec3(m, c, d)

d = cth column of M.

PFGET_MAT_COLVEC3

pfCopyMat(d, m)

D = M.

PFCOPY_MAT

pfAddMat(d, m1, m2)

D = M1 + M2.

none

pfSubMat(d, m1, m2)

D = M1 - M2.

none

pfMultMat(d, m1, m2)

D = M1M2.

none

pfPostMultMat(d, m)

D = DM.

none

pfPreMultMat(d, m)

D = MD.

none

pfTransposeMat(d, m)

D = MT.

none

pfPreTransMat(d, m, x, y, z)

D = TM, where T translates by
(x, y, z).

none

007-1680-080

Matrix Operations

Table 19-2

Routines for 4x4 Matrices (continued)

Routine

Effect

Macro Equivalent

pfPostTransMat(d, x, y, z, m)

D = MT, where T translates by
(x, y, z).

none

pfPreRotMat(d, deg, x, y, z, m)

D = RM, where R rotates by deg
around (x, y, z).

none

pfPostRotMat(d, m, deg, x, y, z)

D = MR, where R rotates by deg
around (x, y, z).

none

pfPreScaleMat(d, x, y, z, m)

D = SM, where S scales by (x, y,
z).

none

pfPostScaleMat(d, m, x, y, z)

D = MS, where S scales by (x, y,
z).

none

pfInvertFullMat(d, m))

D = M-1 for general matrices.

none

pfInvertAffMat(d, m)

D = M-1 with M affine.

none

pfInvertOrthoMat(d, m)

D = M-1 with M orthogonal.

none

pfInvertOrthoNMat(d, m)

D = M-1 with M orthonormal.

none

pfInvertIdentMat(d, m)

D = M-1 with M equal to the
identity matrix.

none

pfEqualMat(d, m)

Returns TRUE if D = M and
FALSE, otherwise

PFEQUAL_MAT

pfAlmostEqualMat(d, m, tol)

Returns TRUE if each element of PFALMOST_EQUAL_MAT
D is within tol of the
corresponding element of M and
FALSE, otherwise

Some of the math routines that take a matrix as an argument are restricted to affine,
orthogonal, or orthonormal matrices; these restrictions are noted by Aff, Ortho, and
OrthoN, respectively. (If such a restriction is not noted in a libpr routine name, the
routine can take a general matrix.)
An affine transformation is one that leaves the homogeneous coordinate unchanged—
that is, in which the last column is (0,0,0,1). An orthogonal transformation is one that
preserves angles. It can include translation, rotation, and uniform scaling, but no

007-1680-080

615

19: Math Routines

shearing or nonuniform scaling. An orthonormal transformation is an orthogonal
transformation that preserves distances; that is, one that contains no scaling.
In the visual simulation library, libpf, most routines require the matrix to be
orthogonal, although this is not noted in the routine names.
The standard order of transformations for a hierarchical scene involves post-multiplying
the transformation matrix for a child by the matrix for the parent. For instance, assume
your scene involves a hand attached to an arm attached to a body. To get a transformation
matrix H for the hand, post-multiply the arm’s transformation matrix (A) by the body’s
(B): H = AB. To transform the hand object (at location h in hand coordinates) to body
coordinates, calculate h’ = hH.
Example 19-1

Matrix and Vector Math Examples

/*
* test Rot of v1 onto v2
*/
{
pfVec3 v1, v2, v3;
pfMatrix m1;
MakeRandomVec3(v1);
MakeRandomVec3(v2);
pfNormalizeVec3(v1);
pfNormalizeVec3(v2);
pfMakeVecRotVecMat(m1, v1, v2);
pfXformVec3(v3, v1, m1);
AssertEqVec3(v3, v2, “Arb Rot To”);
}
/*
* test inversion of Affine Matrix
*/
{
pfVec3 v1, v2, v3;
pfMatrix m1, m2, m3;
MakeRandomVec3(v3);
pfMakeScaleMat(m2, v3[0], v3[1], v3[2]);
pfPreMultMat(m1, m2);
MakeRandomVec3(v1);
pfNormalizeVec3(v1);

616

007-1680-080

Quaternion Operations

MakeRandomVec3(v2);
pfNormalizeVec3(v2);
pfMakeVecRotVecMat(m1, v1, v2);
s = pfLengthVec3(v2)/pfLengthVec3(v1);
pfPreScaleMat(m1, s, s, s, m1);
MakeRandomVec3(v1);
pfNormalizeVec3(v1);
MakeRandomVec3(v2);
pfNormalizeVec3(v2);
pfMakeVecRotVecMat(m2, v1, v2);
MakeRandomVec3(v3);
pfMakeTransMat(m2, v3[0], v3[1], v3[1]);
pfPreMultMat(m1, m2);
pfInvertAffMat(m3, m1);
pfPostMultMat(m3, m1);
AssertEqMat(m3, ident, “affine inverse”);

Quaternion Operations
A pfQuat is the OpenGL Performer data structure (a pfVec4) whose floating point values
represent the components of a quaternion. Quaternions have many beneficial properties.
The most relevant of these is their ability to represent 3D rotations in a manner that
allows relatively simple yet meaningful interpolation between rotations. Much like
multiplying two matrices, multiplying two quaternions results in the concatenation of
the transformations. For more information on quaternions, see the article by Sir William
Rowan Hamilton “On quaternions; or on a new system of imaginaries in algebra,” in
Philosophical Magazine, xxv, pp. 10-13 (July 1844), or refer to the sources noted in the
pfQuat(3pf) man page.
The properties of spherical linear interpolation makes quaternions much better suited
than matrices for interpolating transformation values from keyframes in animations. The
most common usage then is to use pfSlerpQuat() to interpolate between two
quaternions representing two rotational transformations. The quaternion that results
from the interpolation can then be passed to pfMakeQuatMat() to generate a matrix for
use in a subsequent OpenGL Performer call such as pfDCSMat(). While converting a
quaternion to a matrix is relatively efficient, converting a matrix to a quaternion with
pfGetOrthoMatQuat() is expensive and should be avoided when possible.

007-1680-080

617

19: Math Routines

Because a pfQuat is also a pfVec4, all of the pfVec4 routines and macros may be used on
pfQuats as well.
Table 19-3

Routines for Quaternions

Routine

Effect

pfMakeRotQuat(q, a, x, y, z)

Sets q to rotation of a degrees about none
(x, y, z).

pfGetQuatRot(q, a, x, y, z)

Sets *a to angle and (*x, *y, *z) to axis none
of rotation represented by q.

pfConjQuat(d, q)

d = conjugate of q.

PFCONJ_QUAT

pfLengthQuat(q)

Returns length of q.

PFLENGTH_QUAT

pfMultQuat(d, q1, q2)

d = q1 * q2.

PFMULT_QUAT

pfDivQuat(d, q1, d2)

d = q1 / q1.

PFDIV_QUAT

pfInvertQuat(d, q1)

d = 1 / q1.

pfExpQuat(d, q)

d = exp(q).

none

pfLogQuat(d, q)

d = ln(q).

none

pfSlerpQuat(d, t, q1, q2)

d = interpolation with weight t
between q1 (t=0.0) and q2 (t=1.0).

none

pfSquadQuat(d, t, q1, q2, a, b)

d = quadratic interpolation between none
q1 and q2.

pfQuatMeanTangent(d, q1, q2, q3)

d = mean tangent of q1, q2 and q3.

Example 19-2

Macro Equivalent

none

Quaternion Example

/*
* test quaternion slerp
*/
pfQuat q1, q2, q3;
pfMatrix m1, m2, m3, m3q;
pfVec3 axis;
float angle1, angle2, angle, t;

MakeRandomVec3(axis);

618

007-1680-080

Matrix Stack Operations

pfNormalizeVec3(axis);
angle1 = -drand48()*90.0f;
angle2 = drand48()*90.0f;
t = drand48();
pfMakeRotMat(m1, angle1, axis[0], axis[1], axis[2]);
pfMakeRotQuat(q1, angle1, axis[0], axis[1], axis[2]);
pfMakeQuatMat(m3q, q1);
pfMakeRotMat(m2, angle2, axis[0], axis[1], axis[2]);
pfMakeRotQuat(q2, angle2, axis[0], axis[1], axis[2]);
pfMakeQuatMat(m3q, q2);
AssertEqMat(m2, m3q, “make rot quat q2”);
angle = (1.0f-t) * angle1 + t * angle2;
pfMakeRotMat(m3, angle, axis[0], axis[1], axis[2]);
pfMakeRotQuat(q1, angle1, axis[0], axis[1], axis[2]);
pfMakeRotQuat(q2, angle2, axis[0], axis[1], axis[2]);
pfSlerpQuat(q3, t, q1, q2);
pfMakeQuatMat(m3q, q3);
AssertEqMat(m3q, m3, “quaternion slerp”);
{

Matrix Stack Operations
OpenGL Performer allows you to create a stack of transformation matrices, which is
called a pfMatStack.
Table 19-4 lists and describes the matrix stack routines. Note that none of these routines
has a macro equivalent. The matrix at the top of the matrix stack is denoted “TOS,” for
“Top of Stack.”
Table 19-4

007-1680-080

Matrix Stack Routines

Routine

Operation

pfNewMStack()

Allocate storage.

pfResetMStack()

Reset the stack.

pfPushMStack()

Duplicate the TOS and push it on the stack.

619

19: Math Routines

Table 19-4

Matrix Stack Routines (continued)

Routine

Operation

pfPopMStack()

Pop the stack.

pfPreMultMStack()

Premultiply the TOS by a matrix.

pfPostMultMStack()

Postmultiply the TOS by a matrix.

pfLoadMStack()

Set the TOS matrix.

pfGetMStack()

Get the TOS matrix.

pfGetMStackTop()

Get a pointer to the TOS matrix.

pfGetMStackDepth()

Return the current depth of the stack.

pfPreTransMStack()

Pre-multiply the TOS by a translation.

pfPostTransMStack()

Post-multiply the TOS by a translation.

pfPreRotMStack()

Pre-multiply the TOS by a rotation.

pfPostRotMStack()

Post-multiply the TOS by a rotation.

pfPreScaleMStack()

Pre-multiply the TOS by a scale factor.

pfPostScaleMStack()

Post-multiply the TOS by a scale factor.

Creating and Transforming Volumes
libpr provides a number of volume primitives, including sphere, box, cylinder,
half-space (plane), and frustum. libpf uses the frustum primitive for a view frustum
and uses other volume primitives for bounding volumes:
•

pfNodes use bounding spheres.

•

pfGeoSets use bounding boxes.

•

Segments use bounding cylinders.

Defining a Volume
This section describes how to define geometric volumes.

620

007-1680-080

Creating and Transforming Volumes

Spheres

Spheres are defined by a center and a radius, as shown by the pfSphere structure’s
definition:
typedef struct {
pfVec3 center;
float radius;
} pfSphere;

A point p is in the sphere with center c and radius r if |p - c|< r.
Axially Aligned Boxes

An axially aligned box is defined by its two corners with the smallest and largest values
for each coordinate. Its edges are parallel to the X, Y, and Z axes. It is represented by the
pfBox data structure:
typedef struct {
pfVec3 min;
pfVec3 max;
} pfBox;

A point (x, y, z) is in the box if minx < x < maxx, miny < y < maxy, and minz < z < maxz, .
Cylinders

A cylinder is defined by its center, radius, axis, and half-length, as shown by the
definition of the pfCylinder data structure:
typedef struct {
pfVec3 center;
float radius;
pfVec3 axis;
float halfLength;
} pfCylinder;

A point p is in the cylinder with center c, radius r, axis a, and half-length h, if |(p - c) . a|
< h and | (p - c) - ((p - c) . a) a | < r.

007-1680-080

621

19: Math Routines

Half-spaces (Planes)

A half-space is defined by a plane with a normal pointing away from the interior. It is
represented by the pfPlane data structure:
typedef struct {
pfVec3 normal;
float offset;
} pfPlane;

A point p is in the half-space with normal n and offset d if p . n < d.
Frusta

Unlike the other volumes, a pfFrustum is not an exposed structure. You can allocate
storage for a pfFrustum using pfNewFrust() and you can set the frustum using
pfMakePerspFrust() or pfMakeOrthoFrust().

Creating Bounding Volumes
The easiest and most efficient way to create a volume is to use one of the bounding
operations. The routines in Table 19-5 create a bounding volume that encloses other
geometric objects.
Table 19-5

622

Routines to Create Bounding Volumes

Routine

Bounding Volume

pfBoxAroundPts()

Box enclosing a set of points

pfBoxAroundBoxes()

Box enclosing a set of boxes

pfBoxAroundSpheres()

Box enclosing a set of spheres

pfCylAroundSegs()

Cylinder around a set of segments

pfSphereAroundPts()

Sphere around a set of points

pfSphereAroundBoxes ()

Sphere around a set of boxes

pfSphereAroundSpheres ()

Sphere around a set of spheres

007-1680-080

Creating and Transforming Volumes

Bounding volumes can also be defined by extending existing volumes, but in many cases
the tightness of the bounds created through a series of extend operations is substantially
inferior to that of the bounds created with a single pf*Around*() operation.
Table 19-6 lists and describes the routines for extending bounding volumes.
Table 19-6

Routines to Extend Bounding Volumes

Routine

Operation

pfBoxExtendByPt()

Extend a box to enclose a point.

pfBoxExtendByBox()

Extend a box to enclose another box.

pfSphereExtendByPt ()

Extend a sphere to enclose a point.

pfSphereExtendBySphere ()

Extend a sphere to enclose a sphere.

Transforming Bounding Volumes
Transforming the volumes with an orthonormal transformation—that is, with no skew
or nonuniform scaling, is straightforward for all of the volumes except for the axially
aligned box. A straight transformation of the vertices does not suffice because the new
box would no longer be axially aligned; so, an aligned box must be created that encloses
the transformed vertices. Hence, a transformation of a box is not generally reversed by
applying the inverse transformation to the new box.
Table 19-7 lists and describes the routines that transform bounding volumes.
Table 19-7

007-1680-080

Routines to Transform Bounding Volumes

Routine

Operation

pfOrthoXformPlane()

Transform a plane or half-space.

pfOrthoXformFrust()

Transform a frustum.

pfXformBox()

Transform and extend a bounding box.

pfOrthoXformCyl()

Transform a cylinder.

pfOrthoXformSphere()

Transform a sphere.

623

19: Math Routines

Intersecting Volumes
OpenGL Performer provides a number of routines that test for intersection with
volumes.

Point-Volume Intersection Tests
The point-volume intersection test returns PFIS_TRUE if the specified point is in the
volume and PFIS_FALSE otherwise. Table 19-8 lists and describes the routines that test a
point for inclusion within a bounding volume.
Table 19-8

Testing Points for Inclusion in a Bounding Volume

Routine

Test

pfBoxContainsPt()

Point inside a box

pfSphereContainsPt()

Point inside a sphere

pfCylContainsPt()

Point inside a cylinder

pfHalfSpaceContainsPt()

Point inside a half-space

pfFrustContainsPt()

Point inside a frustum

Volume-Volume Intersection Tests
OpenGL Performer provides a number of volume-volume tests that are used internally
for bounding-volume tests when culling to a view frustum or when testing a group of
line segments against geometry in a scene (see “Intersecting with pfGeoSets” on
page 628). You can intersect spheres, boxes, and cylinders against half-spaces and against
frustums for culling. You can intersect cylinders against spheres for testing grouped
segments against bounding volumes in a scene.

624

007-1680-080

Intersecting Volumes

Table 19-9 lists and describes the routines that test for volume intersections.
Table 19-9

Testing Volume Intersections

Routine

Action: Test if A Inside B

pfHalfSpaceContainsSphere()

Sphere inside a half-space

pfFrustContainsSphere()

Sphere inside a frustum

pfSphereContainsSphere()

Sphere inside a sphere

pfSphereContainsCyl()

Cylinder inside a sphere

pfHalfSpaceContainsCyl()

Cylinder inside a half-space

pfFrustContainsCyl()

Cylinder inside a frustum

pfHalfSpaceContainsBox()

Box inside a half-space

pfFrustContainsBox()

Box inside a frustum

pfBoxContainsBox()

Box inside a box

The volume-volume intersection tests are designed to quickly locate empty intersections
for rejection during a cull. If the complete intersection test is too time-consuming, the
result PFIS_MAYBE is returned to indicate that the two volumes might intersect.
The returned value is a bitwise OR of tokens, as shown in Table 19-10.
Table 19-10

Intersection Results

Test Result

Meaning

PFIS_FALSE

No intersection.

PFIS_MAYBE

Possible intersection.

PFIS_MAYBE | PFIS_TRUE

A contains at least part of B.

PFIS_MAYBE | PFIS_TRUE | PFIS_ALL_IN

A contains all of B.

This arrangement allows simple code such as that shown in Example 19-3.

007-1680-080

625

19: Math Routines

Example 19-3

Quick Sphere Culling Against a Set of Half-Spaces

long
HSSContainsSphere(pfPlane **hs, pfSphere *sph, long numHS)
{
long i, isect;
isect = ~0;
for (i = 0 ; i < numHS ; i++)
{
isect &= pfHalfSpaceContainsSphere(sph,hs[i]);
if (isect == PFIS_FALSE)
return isect;
}
/* if not ALL_IN all half spaces, don’t know for sure */
if (!(isect & PFIS_ALL_IN))
isect &= ~PFIS_TRUE;
return isect;
}

Creating and Working with Line Segments
A pfSeg represents a line segment starting at position pos and extending for a distance
length in the direction dir:
typedef struct {
pfVec3 pos;
pfVec3 dir;
float length;
} pfSeg;

The routines that operate on pfSegs assume that dir is of unit length and that length is
positive; otherwise, the results of operations are undefined.
You can create line segments in four ways:

626

•

Specify a point and a direction directly in the structure—pfSeg().

•

Specify two endpoints— pfMakePtsSeg().

•

Specify one endpoint and an orientation in polar coordinates—pfMakePolarSeg().

•

Specify starting and ending distances along an existing segment—pfClipSeg().

007-1680-080

Creating and Working with Line Segments

Intersection tests are the most important operations that use line segments. You can test
the intersection of segments with volumes (half-spaces, spheres, and boxes), with 2D
geometry (planes and triangles), and with geometry inside pfGeoSets.

Intersecting with Volumes
OpenGL Performer supports intersections of segments with three types of convex
volumes. pfHalfSpaceIsectSeg() intersects a segment with the half-space defined by a
plane with an outward facing normal. pfSphereIsectSeg() intersects with a sphere and
pfBoxIsectSeg() intersects with an axially aligned box.
The intersection test of a segment and a convex volume can have one of five results:
•

The segment lies entirely outside the volume.

•

The segment lies entirely within the volume.

•

The segment lies partially inside the volume with the starting point inside.

•

The segment lies partially inside the volume with the ending point inside.

•

The segment lies partially inside the volume with both endpoints outside.

As with the volume-volume tests, the segment-volume intersection routines return a
value that is the bitwise OR of some combination of the tokens PFIS_TRUE,
PFIS_ALL_IN, PFIS_START_IN, and PFIS_MAYBE. (When PFIS_TRUE is set
PFIS_MAYBE is also set for consistency with those routines that do quick intersection
tests for culling.)
The functions take two arguments that return the distances along the segment of the
starting and ending points. The return values are designed so that you can AND them
together for testing for the intersection of a segment against the intersection of a number
of volumes. For example, a convex polyhedron is defined as the intersection of a set of
half-spaces. Example 19-4 shows how to intersect a segment with a polyhedron.
Example 19-4

Intersecting a Segment With a Convex Polyhedron

long
HSSIsectSeg(pfPlane **hs, pfSeg *seg, long nhs, float *d1,
float *d2)
{
long retval = 0xffff;
for (long i = 0 ; i < nhs ; i++)
{

007-1680-080

627

19: Math Routines

retval &= pfHalfSpaceIsectSeg(hs[i], seg, d1, d2);
if (retval == 0)
return 0;
pfClipSeg(seg, *d1, *d2);
}
return retval;
}

Note that these routines do not actually clip the segment. If you want the segment to be
clipped to the interior of the volume, you must call pfClipSeg(), as in the example above.

Intersecting with Planes and Triangles
Intersections with planes and triangles are simpler than those with volumes.
pfPlaneIsectSeg() and pfTriIsectSeg() return either PFIS_TRUE or PFIS_FALSE,
depending on whether an intersection has occurred. The distance of the intersection
along the segment is returned in one of the arguments.

Intersecting with pfGeoSets
You can intersect line segments with the drawable geometry that is within pfGeoSets by
calling pfGSetIsectSegs(). The operation is very similar to that of pfNodeIsectSegs(),
except that rather than operating on an entire scene graph, only the triangles within the
pfGeoSet are “traversed.”
pfGSetIsectSegs() takes a pfSegSet and tests to see whether any of the segments intersect
the polygons inside the specified pfGeoSet. By default, information about the closest
intersection along each segment is returned as a set of pfHit objects, one for each line
segment in the request. Each pfHit object indicates the location of the intersection, the
normal, and what element was hit. This element identification includes the index of the
primitive within the pfGeoSet and the triangle index within the primitive (for tristrips
and quads primitives), as well as the actual triangle vertices.
You can also extract information from a pfHit object using pfQueryHit() and
pfMQueryHit(). (See “Intersection Requests: pfSegSets” and “Intersection Return Data:
pfHit Objects” in Chapter 4 for more information about pfSegSets and pfHit objects.) The
principal difference between those routines and pfGSetIsectSegs() is that with
pfGSetIsectSegs() information concerning the libpf scene graph (such as
transformation, geode, name, and path) is never used.

628

007-1680-080

Creating and Working with Line Segments

Two types of intersection testing are possible, as shown in Table 19-11.
Table 19-11

Available Intersection Tests

Test Name

Function

PFTRAV_IS_GSET

Intersect the segment with the bounding box of the pfGeoSet.

PFTRAV_IS_PRIM

Intersect the segment with the polygon-based primitives inside the
pfGeoSet.

You can use PFTRAV_IS_GSET for crude collision detection and PFTRAV_IS_PRIM for
fine-grained testing. You can enable both bits and dynamically choose whether to go
down to the primitive level by using a discriminator callback (see “Discriminator
Callbacks”). pfGSetIsectSegs() performs only primitive-level testing for pfGeoSets
consisting of triangles (PFGS_TRIS), quads (PFGS_QUADS), and tristrips
(PFGS_TRISTRIPS), and all are decomposed into triangles.
Intersection Masks

Each pfGeoSet has an intersection mask that you set using pfGSetIsectMask(). The mask
in the pfGeoSet is useful when pfGeoSets are embedded in a larger data structure; it
allows you to define pfGeoSets to belong to different classes of geometry for
intersection—for example, water, ground, foliage. pfGSetIsectSegs() also takes a mask,
and an intersection test is performed only if the bitwise AND of the two masks is
nonzero.
Discriminator Callbacks

If a callback is specified in pfGSetIsectSegs(), the callback function is invoked when a
successful intersection occurs, either with the bounding box of the pfGeoSet or with a
primitive. The discriminator can decide what action to take based on the information
about the intersection contained in a pfHit object. The return value from the
discriminator determines whether the current intersection is valid and should be copied
into the return structure, whether the rest of the geometry in the pfGeoSet is examined,
and whether the segment should be clipped before continuing.
Unless the return value includes the bit PFTRAV_IS_IGNORE, the intersection is
considered successful and is copied into the array of pfHit structures for return.

007-1680-080

629

19: Math Routines

The bits of the PFTRAV_* tokens determine whether to continue, as shown in
Table 19-12.
Table 19-12

Discriminator Return Values

Result

Meaning

PFTRAV_CONT

Continue examining geometry inside the pfGeoSet.

PFTRAV_PRUNE

Terminate the traversal now.

PFTRAV_TERM

Terminate the traversal now.

The bits PFTRAV_IS_CLIP_END and PFTRAV_IS_CLIP_START cause the segment to be
clipped at the end or at the start using the intersection point. By default, in the absence
of a discriminator, segments are end-clipped at each successful intersection at the finest
level (bounding box or primitive level) requested. Hence, the closest intersection point is
always returned.
The discriminator is passed a pfHit. You can use pfQueryHit() to examine information
about the intersection, including which segment number within the pfSegSet the
intersection is for and the current segment as clipped by previous intersections.

General Math Routine Example Program
Example 19-5 demonstrates the use of many of the available OpenGL Performer math
routines.
Example 19-5

Intersection Routines in Action

/*
* simple test of pfCylIsectSeg
*/
{
pfVec3 tmpvec;
pfSetVec3(pt1, -2.0f, 0.0f, 0.0f);
pfSetVec3(pt2, 2.0f, 0.0f, 0.0f);
pfMakePtsSeg(&seg1, pt1, pt2);
pfSetVec3(cyl1.axis, 1.0f, 0.0f, 0.0f);
pfSetVec3(cyl1.center, 0.0f, 0.0f, 0.0f);

630

007-1680-080

General Math Routine Example Program

cyl1.radius = 0.5f;
cyl1.halfLength = 1.0f;
isect = pfCylIsectSeg(&cyl1, &seg1, &t1, &t2);
pfClipSeg(&clipSeg, &seg1, t1, t2);
AssertFloatEq(clipSeg.length, 2.0f, “clipSeg.length”);
pfSetVec3(tmpvec, 1.0f, 0.0f, 0.0f);
AssertVec3Eq(clipSeg.dir, tmpvec, “clipSeg.dir”);
pfSetVec3(tmpvec, -1.0f, 0.0f, 0.0f);
AssertVec3Eq(clipSeg.pos, tmpvec, “clipSeg.pos”);
}
/*
* simple test of pfTriIsectSeg
*/
{
pfVec3 tr1, tr2, tr3;
pfSeg seg;
float d = 0.0f;
long i;
for (i = 0 ; i < 30 ; i++)
{
float alpha = 2.0f * drand48() - 0.5f;
float beta = 2.0f * drand48() - 0.5f;
float lscale = 2.0f * drand48();
float target;
long shouldisect;
MakeRandomVec3(tr1);
MakeRandomVec3(tr2);
MakeRandomVec3(tr3);
MakeRandomVec3(pt1);
pfCombineVec3(pt2, alpha, tr2, beta, tr3);
pfCombineVec3(pt2, 1.0f, pt2, 1.0f - alpha - beta, tr1);
pfMakePtsSeg(&seg, pt1, pt2);
target = seg.length;
seg.length = lscale * seg.length;
isect = pfTriIsectSeg(tr1, tr2, tr3, &seg, &d);
shouldisect = (alpha >= 0.0f &&
beta >= 0.0f &&
alpha + beta <= 1.0f &&
lscale >= 1.0f);

007-1680-080

631

19: Math Routines

if (shouldisect)
if (!isect)
printf(“ERROR: missed\n”);
else
AssertFloatEq(d, target, “hit at wrong distance”);
else if (isect)
printf(“ERROR: hit\n”);
}
/*
* simple test of pfCylContainsPt
*/
{
pfCylinder cyl;
pfVec3 pt;
pfVec3 perp;
pfSetVec3(cyl.center, 1.0f, 10.0f, 5.0f);
pfSetVec3(cyl.axis, 0.0f, 0.0f, 1.0f);
pfSetVec3(perp, 1.0f, 0.0f, 0.0f);
cyl.halfLength = 2.0f;
cyl.radius = 0.5f;
pfCopyVec3(pt, cyl.center);
if (!pfCylContainsPt(&cyl, pt))
printf(“center of cylinder not in cylinder!!!!\n”);
pfAddScaledVec3(pt, cyl.center, 0.9f*cyl.halfLength,
cyl.axis);
if (!pfCylContainsPt(&cyl, pt))
printf(“0.9*halfLength not in cylinder!!!!\n”);
pfAddScaledVec3(pt, cyl.center, -0.9f*cyl.halfLength,
cyl.axis);
if (!pfCylContainsPt(&cyl, pt))
printf(“-0.9*halfLength not in cylinder!!!!\n”);
pfAddScaledVec3(pt, cyl.center, -0.9f*cyl.halfLength,
cyl.axis);
pfAddScaledVec3(pt, pt, 0.9f*cyl.radius, perp);
if (!pfCylContainsPt(&cyl, pt))
printf(printf(“-0.9*halfLength not in cylinder!!\n”);
pfAddScaledVec3(pt, cyl.center, 0.9f*cyl.halfLength,
cyl.axis);

632

007-1680-080

General Math Routine Example Program

pfAddScaledVec3(pt, pt, -0.9f*cyl.radius, perp);
if (!pfCylContainsPt(&cyl, pt))
printf(“-0.9*halfLength not in cylinder!!!!\n”);
pfAddScaledVec3(pt, cyl.center, 1.1f*cyl.halfLength,
cyl.axis);
if (pfCylContainsPt(&cyl, pt))
printf(“1.1*halfLength in cylinder!!!!\n”);
pfAddScaledVec3(pt, cyl.center, -1.1f*cyl.halfLength,
cyl.axis);
if (pfCylContainsPt(&cyl, pt))
printf(“-1.1*halfLength in cylinder!!!!\n”);
pfAddScaledVec3(pt, cyl.center, -0.9f*cyl.halfLength,
cyl.axis);
pfAddScaledVec3(pt, pt, 1.1f*cyl.radius, perp);
if (pfCylContainsPt(&cyl, pt))
printf(“1.1*radius in cylinder!!!!\n”);
pfAddScaledVec3(pt, cyl.center, 0.9f*cyl.halfLength,
cyl.axis);
pfAddScaledVec3(pt, pt, -1.1f*cyl.radius, perp);
if (pfCylContainsPt(&cyl, pt))
printf(“1.1*radius in cylinder!!!!\n”);
}

007-1680-080

633

Chapter 20

20. Statistics

This chapter describes the OpenGL Performer profiling utilities. Statistics are available
on nearly every aspect of OpenGL Performer’s operation and can be used to diagnose
both functionality and performance problems, as well as for writing benchmarks and for
load management. For more detailed information on interpreting statistics to tune the
performance of your application, refer to Chapter 21, “Performance Tuning and
Debugging.”
To collect most OpenGL Performer statistics, all you have to do is enable them; OpenGL
Performer then collects them automatically for you in pfStats and pfFrameStats data
structures (for libpr and libpf, respectively). You can query the contents of these
structures from your program, or write the data to files. A libpf application can also
display the contents of a pfFrameStats structure in a channel by calling
pfDrawChanStats() or pfDrawFStats(). The statistics drawn for a channel are the
statistics accumulated in the channel’s own pfFrameStats. Such a display is not necessary
for statistics collection. The pointer to the pfFrameStats structure for a channel can be
obtained with pfGetChanFStats(). You can then control which statistics for the channel
are being accumulated.
Most of the OpenGL Performer demo programs display some subset of these statistics.
This chapter first explains some of the complex graphical displays and then discusses
how to display statistics from a libpf-based application. Subsequent sections explain
how to access and manipulate statistics from within an application. Topics include
enabling and disabling statistics classes, printing, querying, and copying statistics data,
as well as some basic examples showing common uses of statistics. At the end of this
chapter is a discussion of the different statistics classes for libpr and libpf along with
details of their use.

007-1680-080

635

20: Statistics

Interpreting Statistics Displays
Many types of statistics can be displayed in a channel. Most such displays consist simply
of labeled numbers and are fairly self-explanatory; however, some of the displays, such
as the stage timing graph, warrant further explanation.
OpenGL Performer tracks the time spent in the application, cull, and draw stages of the
rendering pipeline. The basic statistics display shows a timing graph for each stage of the
past several frames, as well as showing the current frame rate and load information. This
profiling diagram is useful for optimizing both the database and application structure.
Figure 20-1 shows a sample stage timing graph from an OpenGL Performer demo
program. It might be helpful to refer to a running example as well—by turning on a
statistics display in perfly, for instance—while reading this section.

Figure 20-1

Stage Timing Statistics Display

The statistics diagram in Figure 20-1 is the simplest of the standard statistics displays.
There are several other standard display formats, each emphasizing other classes of
statistics. Statistics collection, though highly optimized, can take extra time in OpenGL
Performer operations. Because of this, you have a great deal of fine control over exactly
what statistics are currently being collected and what statistics are being displayed.
Statistics are divided into classes (separated into vertically stacked boxes in a display),
and into modes within each class. The next several sections describe the classes shown in
a typical statistics display.

Status Line
The top line of a standard statistics display, above the box that contains the rest of the
statistics, shows the current average frame rate followed by a slash and the target frame
rate. (To set a target frame rate, call pfFrameRate().) The rest of that status line indicates
what frame-rate control method you are using (FLOAT, FREE, LIMIT, or LOCK—for

636

007-1680-080

Interpreting Statistics Displays

details, see “Achieving the Frame Rate” in Chapter 5), your multiprocess model (set with
pfMultiprocess()), and the average time (in milliseconds) spent in the channel draw
callback. An optional part of the status line indicates the number of triangles in the scene.

Stage Timing Graph
The main part of the timing display is the stage timing graph, occupying the top portion
of the statistics display. The red vertical lines (the darker ones in Figure 20-1) mark video
retrace intervals, which occur at the video refresh rate of the system (commonly 60 times
per second); a field is the period of time between two video retrace boundaries. The
green vertical lines (the lighter ones in the figure) indicate frame boundaries. Note that
frame boundaries are always on field boundaries and are an integral number of fields.
The segmented horizontal line segments in the top portion of the timing graph show the
time taken by each of the OpenGL Performer pipeline stages and additional processes:
i (intersection), a (application), c (cull), d (draw), l (lpoint), db (dbase), and cx (compute)
for each of the four frames shown (0 through 3). On screen, all stages belonging to a given
frame are drawn in one color; different colors indicate different frames. You will notice
that the application lines show a change in color. This point is where pfFrame() returned
and is the start of the next application frame. At that point is a label for the stage name
and the age of the frame being represented. The stages of the most recent frame, at the
right of the graph, are marked a0, c0, and d0; previous frames have higher numbers (so
“a-1” indicates the application stage of the immediately previous frame).
All stages performed by the same process are connected by vertical lines. If two stages
are performed by different processes, they are not connected by a vertical line. In most
multiprocessing modes, a stage of one frame occurs at the same time as another stage for
a different frame, so that (for instance) d0 is directly below c-1 in the graph. The exception
is the PFMP_CULLoDRAW model, in which the cull and draw stages for a given frame
are performed in tandem; in this mode d0 is directly below c0 in the graph. (In
Figure 20-1, the PFMP_APPCULLDRAW model was used and all stages are part of
separate processes.)
These stage timings are helpful when choosing a process model and balancing the cull
and draw tasks for a database. Furthermore, the timing graph can show you how close
you are to an improvement in frame rate as you view the database.
The timing lines for each stage are broken into pieces displayed at slightly different
heights and thicknesses to show the time taken by significant subtasks within each stage.

007-1680-080

637

20: Statistics

Raised segments reflect time spent in user code, intermediate lines reflect time spent in
OpenGL Performer code, and lowered lines reflect time waiting on other operations.
Figure 20-2 illustrates the parts of a draw-stage timing line. Note that this figure is not
drawn to scale; sizes are exaggerated in order to discuss the individual parts more easily.

Before pfDraw
During pfDraw
After pfDraw
Drawing stats

Figure 20-2

Conceptual Diagram of a Draw-Stage Timing Line

The following explains the potential displayed elements of each stage line in the timing
graph. Note that if you run your application or perfly, you may not see some parts if
the corresponding operations are not needed.
The application stage is divided into five subsegments, starting at the point where
pfFrame() returns and the new application frame is beginning:

638

•

The time spent in the application’s main loop between the pfFrame() call and the
pfSync() call (highest segment in application line, drawn as a thick, bright line).

•

The time spent cleaning the scene graph from application changes during pfSync();
drawn as mid-height thick, bright line. This will also include the time for
pfAppFrame(),which is called from pfSync() if not already called for the current
frame by the user. pfSequences are also evaluated as part of pfAppFrame().

•

The time spent sleeping in pfSync() while waiting for the next field or frame
boundary (depending on pfPhase and process model); the lowest point in the
application line, drawn as a thin pale dotted line. Note that in single process with
pfPhase of FREE_RUN, there will be a sleep period to wait for the swapbuffer of the
draw to complete before continuing with the application since any other graphics
call would effectively force such a sleep anyway and in a place where its timing
effect could not be measured.

•

The time spent in the application code between calling pfSync() and calling
pfFrame(); drawn as bright raised line. This is the critical path section and this line
should be as small as possible or non-existent.

007-1680-080

Interpreting Statistics Displays

•

The time spent in pfFrame() cleaning the scene graph after any changes that might
have been made in the previous subsegment, and then checking intersections;
drawn as mid-height thick bright line. This line should typically be very small or
non-existent as it is part of the critical path and implies database changes between
pfSync() and pfFrame() that would be an expensive place to do such changes.

•

The time spent waiting while the cull and other downstream process copy updated
information from the application and then starting the downstream stages on the
now-finished frame (drawn as a low thin line). The end of this line is where
pfFrame() returns and the user main application section (or post frame section)
starts again.

The cull stage is divided into only two subsegments:
•

The time spent receiving updates from the application (in some multiprocessing
models, this overlaps with the last subsegment of the application stage). This time is
displayed for all channels even though it is only done one time. This is drawn as a
lowered thick line.

•

The time spent in the channel cull callback for the given channel, including time
spent in pfCull() (drawn as raised line). Note that there may be a large space
between this and the update line if there are multiple channels on the same pfPipe
that are processed first.

The draw stage has potentially six subsegments:

007-1680-080

•

The time spent in the channel draw callback before the call to pfDraw() (a very
short thick dark raised segment). This includes the time for your call to
pfClearChan(). However, under normal circumstances, this segment should barely
be visible at all. Operations taking place during this time should only be
latency-critical since they are holding off the draw for the current frame.

•

The time spent by OpenGL Performer traversing the scene graph in pfDraw(); it is
drawn as lowered bright thick segment. This should typically be the largest
segment of the draw line.

•

The time spent in the channel draw callback after pfDraw() (another short thick
dark raised segment). On InfiniteReality, if graphics pipeline timing statistics have
been enabled (PFFSTATS_ENGFXPFTIMES), this line will include the time to finish
the fill for this channel. Otherwise, it only includes the time for the CPU to execute
and send graphics commands, and graphics pipeline processing from this channel
could impact the timing of other channels.

639

20: Statistics

•

The time to rendering raster light points computed by a forked lpoint process. This
is drawn as a very raised bright line and if it exists will be the highest point in the
draw line.

•

The last channel drawn on the pipe will include the time for the graphics pipeline to
finish its drawing. Even if you have no operations after pfDraw() in your draw
callback, this line for the last channel might look quite long particularly if you are
very fill-limited and do not have InfiniteReality graphics pipeline statistics enabled.
It is possible for rendering calls issued in the previous section to fill up the graphics
FIFO and have calls issued on this section have to wait while the graphics pipeline
processes the commands and FIFO drains, making the time look longer than
expected. If there is no forked lpoint process, this line will be combined with the
post-draw line of the last pfChannel.

•

The time spent waiting for the graphics pipeline to finish drawing the current
frame, draw the channel statistics (for all channels), and make the call to swap color
buffers. This is drawn as a pale dotted line. The hardware will complete the
swapbuffers upon the following vertical field or frame line.

Below the stage timing lines, the average time for each stage (in milliseconds) is shown.
Note that the time given for the draw stage is the same as the time shown for the draw
stage on the status line above the statistics box.

Load and Stress
The lower portion of the channel statistics diagram shows the recent history of graphics
load and stress management. The load measure is based on the amount of time taken to
draw previous frames in the channel relative to the specified goal frame time. A wavy
red horizontal line is drawn to show the last three seconds of graphics load. A pair of
white horizontal lines represent the upper and lower bounds of graphics load for
invoking stress management. Thus, when the red line wanders outside the boundaries
set by the white lines, stress management is invoked.
Stress management causes scaling of LODs in the database to meet the target frame rate
with maximum scene detail. The last three seconds of stress are shown in white while
stress management is running. Thus, the channel statistics graph can be used to tune the
upper and lower bounds of the hysteresis band for invoking stress management and for
tuning LODs of objects in the database.

640

007-1680-080

Interpreting Statistics Displays

CPU Statistics
The CPU statistics keep track of system usage and require that the corresponding
hardware statistics be enabled:
pfEnableStatsHw(PFSTATSHW_ENCPU);

The percentage of time CPUs spend idle, busy, in user code, waiting on the Graphics
Pipeline, or on the swapping of memory is calculated. The statistics packages counts the
number of the following:
•

Context switches (process and graphics)

•

System calls

•

Times the graphics FIFO is found to be full

•

Times a CPU went to sleep waiting on a full graphics FIFO

•

Graphics pipeline IOCTLs issued (by the system)

•

Swapbuffers seen

All of these statistics are computed over an elapsed period of time.
Note: Use an interval of at least one second.
PFSTATSHW_CPU_SYS
This mode calculates the above CPU statistics for the entire system. This
mode is enabled by default.
PFSTATSHW_CPU_IND
This mode calculates the above CPU statistics for each individual CPU;
it is much more expensive than using just the summed statistics.
CPU statistics, illustrated (with some other statistics) in Figure 20-3, give you
information on system usage and load. The numbers shown correspond exactly to
numbers given by osview; they are updated every update period just like other statistics
(see “Setting Update Rate” on page 653 for information on how to change the update
rate). These numbers represent averages (per second) across all CPUs; thus, if one or
more CPUs is busy with some other task, the CPU statistics shown may not accurately
reflect OpenGL Performer CPU use. Note that the top line of the CPU statistics panel
shows the total number of frames during the last update period and the total time
elapsed during that period.

007-1680-080

641

20: Statistics

RTMon Statistics (IRIX Only)

The IRIX kernel collects timestamps using the rtmon daemon, rtmond(1). OpenGL
Performer issues rtmon timestamps for all operations in the timing graph if the rtmon
statistics are enabled with pfStatsClass(PFFSTATS_ENRTMON, PF_ON).

Figure 20-3

642

Other Statistics Classes

007-1680-080

Interpreting Statistics Displays

Rendering Statistics
Several other classes of statistics can be shown, each representing a different aspect of
rendering performance. Some of these classes show the following:
•

Data about visible geometry, including a histogram showing the percentage of
triangles in the scene that are part of strips of a given length (from 1 to 14). Quads
are counted as strips of length two; independent triangles count as strips of length
one. This histogram is mostly useful as a diagnostic to see how well your database
is structured for drawing efficiency; if it shows too many very short strips, you may
want to go back and restructure your database. (As a general rule of thumb,
consider a “very short strip” to be one that is less than four triangles long but that
number may vary depending on your database). To enable these statistics on a
channel do the following:
pfFStatsClass(pfGetChanFStats(chan),
PFSTATS_ENGFX, PFSTATS_ON);

•

A summary of the graphics state operations (including loading of textures) and of
the number of operations that have recently been performed on the transformation
stack (also part of the GFX stats class), the number of libpf nodes being drawn in
several categories (including billboards, light points, and geodes), plus the number
of nodes of each type evaluated in the application and cull stages. The following
enables these statistics:
pfFStatsClass(pfGetChanFStats(chan),
PFSTATS_ENDB, PFSTATS_ON);

•

Cull statistics, including how many nodes and pfGeoSets are being tested, how
many are accepted, and how many are rejected by the libpf culling task. plus the
number of nodes of each type evaluated in the application and cull stages. The
following enables these statistics:
pfFStatsClass(pfGetChanFStats(chan),
PFFSTATS_ENDB, PFSTATS_ON);

•

Graphics pipeline timing statistics showing the time spent rendering as measured
by the graphics pipeline. This timing is then used internally for more accurate load
management. This is supported by InfiniteReality graphics platforms. These
statistics are enabled as follows:
pfFStatsClass(pfGetChanFStats(chan),
PFSTATSHW_ENGFXPIPE_TIMES, PFSTATS_ON);

007-1680-080

643

20: Statistics

Fill Statistics
The fill statistics display indicates how many millions of pixels have been drawn since
the last statistics update. (For information on setting the length of time between statistics
updates, see “Setting Update Rate” on page 653). It also computes the average depth
complexity of the image, which is the average number of times each pixel was touched per
frame.
The depth complexity of a scene is also be displayed in the main channel. Each pixel will
be colored according to how many times that pixel was written to during display, rather
than according to the current rendering modes. The colors used range from dark blue
(not written to at all) to bright pink (written over many times). This color scheme is used
in calculating fill statistics; the coloring is done whenever you gather fill statistics even
when you are not displaying the totals in your channel statistics display.
Stencil planes are used to store the number of times a pixel is written and, thus, to
calculate fill statistics. If n stencil planes are available, no more than 2n writes to any given
pixel will be counted. By default, the calculation of fill statistics uses three stencil planes;
to change that default, call pfStatsHwAttr().
Fill statistics are part of the libpr pfStats statistics but can be enabled on both pfStats
and pfFrameStats classes. To enable fill statistics, simply use the following:
pfStatsClass(statsptr,
PFSTATSHW_ENGFXPIPE_FILL, PFSTATS_ON);

To enable fill statistics for a channel’s pfFrameStats, use the following:
pfFStatsClass(pfGetChanFStats(chan),
PFSTATSHW_ENGFXPIPE_FILL, PFSTATS_ON);

Examples of fill statistics can be found in perfly and in
/usr/share/Performer/src/pguide/libpr/C/fillstats.c for IRIX and
Linux and in %PFROOT%/Src/pguide/libpr/C/fillstats.c for Microsoft
Windows.

Collecting and Accessing Statistics in Your Application
If you just want to bring up a statistics display in your application, you may not need to
know details about the data structures used for statistics. If, however, you want to do
more complicated statistics-handling (including collecting statistics without displaying

644

007-1680-080

Collecting and Accessing Statistics in Your Application

them), you need more advanced information. This section provides a general overview
of statistics manipulation, followed by subsections containing specific information.
If you use libpf, a wide variety of statistics-manipulation functions is available. If you
use libpr, however, you must do some things on your own. For instance, you have to
bind your own pfStats structure in which to accumulate statistics.
Furthermore, you cannot access some kinds of statistics except through libpf calls—for
instance, you cannot get culling statistics using libpr calls. If you want full access to
statistics, you must use libpf. There are, however, libpr routines that allow you to do
your own cumulative totaling and averaging of collected statistics.
To create your own statistics display, enable the statistics classes you want to use and
disable any modes you do not want to use. Then enable any relevant hardware, if
necessary, with pfEnableStatsHw().
To ensure the accuracy of timing with your rendering statistics, you want to flush the
graphics pipeline before calling pfGetTime(). You can do this with glFinish(). These calls
are expensive and should not be done more than at the start and end of drawing in the
frame.

Displaying Statistics Simply
To put up a simple statistics display, all you have to do is call the function
pfDrawChanStats() and pass it a pointer to the pfChannel whose statistics you want to
display. The pfDrawChanStats() routine can be called from any process within the
application; the statistics will be displayed in the channel specified.
If you want to display one channel’s statistics in another channel, call pfDrawFStats();
for an example of this technique, as well as the enabling and disabling of every statistics
class, see the statistics programming example in the file
/usr/share/Performer/src/pguide/libpf/C/stats.c for IRIX and Linux and
in file %PFROOT%/Src/pguide/libpf/C/stats.c for Microsoft Windows.
By default, a statistics display shows all enabled statistics. If you want to show only a
subset of the statistics you are collecting, call pfChanStatsMode() with an enabling
bitmask indicating which classes are to be displayed.

007-1680-080

645

20: Statistics

Enabling and Disabling Statistics for a Channel
For efficiency, you may want to turn off statistics collection for a given channel when you
are not displaying that channel’s statistics. In particular, the stage timing statistics are
enabled by default; so, if you are using a channel whose statistics you do not care about,
you should disable statistics for that channel. To turn off statistics for a channel, use the
following:
pfFStatsClass(pfGetChanFStats(chan),
PFSTATS_ALL, PFSTATS_OFF);

Use the same function with different parameters to enable all or specific classes of
statistics for a channel. You can specify which classes to enable in order to minimize
statistics collection overhead.

Statistics in libpr and libpf—pfStats Versus pfFrameStats
libpf statistics accumulate into a pfFrameStats structure to later be displayed, printed,
queried, or otherwise processed. The pfFrameStats structure actually contains four
buffers of statistics: a buffer for the previous frame’s statistics, a buffer of averaged
statistics for the previous update period, a buffer of accumulated statistics for the current
update period, and a buffer of statistics being accumulated for the current frame.
The pfFrameStats structure is built upon the libpr pfStats structure; so, the
pfFrameStats API includes routines to duplicate the functionality of pfStats. The
duplicated API exists because the routines cannot be intermixed. pfStats routines can
only be used on pfStats structures and pfFrameStats routines can only be called with
pfFrameStats structures. However, pfstats classes and class modes (designated with the
PFSTATS_ prefix) can be enabled on a pfFrameStats structure.
The pfStats statistics classes include the system and hardware statistics for the graphics
pipeline and the CPU, as well as the pixel fill statistics and rendering statistics on
geometry, graphics state, and matrix transformations. Some of the libpr statistics
commands, such as pfEnableStatsHw(PFSTATSHW_ENGFXPIPE_FILL), require an
active graphics context and thus should only be called from the draw process. However,
these commands are usually never necessary in a libpf application because the
pfFrameStats operation will handle these commands automatically.

646

007-1680-080

Collecting and Accessing Statistics in Your Application

Statistics Class Structures

The pfFrameStats structure and the pfStats structure are both inherited from the pfObject
structure. Thus, you can use the pfObject routines (pfCopy(), pfPrint(), pfDelete(),
pfUserData(), pfGetType(), and so on) with pfStats and pfFrameStats structures.
However, some pfObject routines will not support all of the semantics of a pfStats or
pfFrameStats structure; so, some pfStats versions of a few of these routines take extra
arguments. These routines will have a pfFrameStats version as well. In particular,
pfCopyStats() and pfCopyFStats() should be used to copy pfStats and pfFrameStats
structures, respectively.
Routines that have “FStats” in their names (rather than just “Stats”) expect to be passed
a full pfFrameStats structure rather than a pfStats structure. The pfFrameStats API
includes additional routines beyond pfStats for supporting libpf statistics. For
example, pfDrawFStats() displays statistics in a channel and pfFStatsCountNode()
accumulates the static database statistics for the scene graph rooted at the provided node.
Additionally, pfFrameStats has special support for the multiprocessed environment of
libpf and ensures that the statistics operations are all done in the correct process. All
modifying of a pfFrameStats structure, including enabling and disabling of classes,
printing, and copying, should all be done in the application process. pfDrawFStats() and
pfDrawChanStats() can be called in either the application process or the draw process.

Statistics Rules of Use
Enabling and disabling of statistics and setting of modes and attributes on a statistics
structure should always be done in the application process; the settings will
automatically be passed down the process pipeline. To enable classes of statistics on a
pfFrameStats, call pfFStatsClass() and provide a statistics structure, a bitmask of
statistics-enabling tokens (tokens with “STATS_EN” in their names) for the desired
classes, and the token PFSTATS_ON. Obtain the statistics structure from the desired
channel by calling pfGetChanFStats() as follows:
pfFStatsClass(pfGetChanFStats(chan), PFFSTATS_ENCULL |
PFFSTATS_ENDB, PFSTATS_ON);

It enables the cull statistics and database statistics classes, leaving settings alone for any
other classes. Notice that the classes specific to pfFrameStats have a PFFSTATS_ prefix.
You can use PFSTATS_SET instead of PFSTATS_ON to enable only the specified classes
(disabling any others that might already be enabled).

007-1680-080

647

20: Statistics

Statistics Tokens

There are five main types of statistics tokens:
•

Statistics class-enable bitmasks, used for selecting a set of classes to enable with
pfStatsClass(). Class enables and disables are specified with bitmasks. Each
statistics class has an enable token: a PFSTATS_EN* token that can be ORed with
other statistics enable tokens and the result passed in to enable and disable statistics
operations. These bitmasks are also used when printing with pfPrint() or copying
with pfCopy() and pfCopyStats() as well as with the pfResetStats(), pfClearStats(),
pfAverageStats(), and pfAccumulateStats() routines (and their pfFrameStats
counterparts). These tokens are of the form PFSTATS_EN* and PFFSTATS_EN* for
pfStats and pfFrameStats class, respectively. The PFSTATS_ALL token selects all
statistics classes and also all statistics buffers in the case of a pfFrameStats structure.
The token PFSTATS_EN_MASK selects all pfStats classes and the token
PFFSTATS_EN_MASK selects all pfFrameStats statistics classes, which includes all
pfStats classes.

•

Value tokens, used to specify how to set a value for a specified pfStats or
pfFrameStats class enable or mode. Value tokens include PFSTATS_ON,
PFSTATS_OFF, and PFSTATS_DEFAULT. Another value token, PFSTATS_SET, is
used to specify that the entire class enable or mode bitmask should be set to the
specified mask. These tokens are used in conjunction with the class bitmasks and
the class name tokens for pfStatsClass() and pfStatsClassMode().

•

Class name tokens, used to name a specific class. For instance, these tokens can be
passed to pfStatsClassMode() to set individual modes of a statistics class.

•

Class mode tokens, of which each statistics class has its own and which have the
form PFSTATS_class_mode and PFFSTATS_class_mode for pfStats and
pfFrameStats class modes, respectively.

•

Statistics query tokens, used with pfQueryStats(), pfMQueryStats(),
pfQueryFStats(), and pfMQueryFStats(). These tokens are of the form
PFSTATSVAL_* and PFFSTATSVAL_* and have matching pfStatsVal* types for
holding the returned data. The token PFFSTATS_BUF_MASK selects the
pfFrameStats statistics buffers.

Statistics Buffers

You can only access the PREV and CUM statistics buffers from the OpenGL Performer
application process. Statistics from desired buffers in other processes should be queried

648

007-1680-080

Collecting and Accessing Statistics in Your Application

in the application process and then passed down the process pipeline. You can do this
using the channel data utility.
The AVG buffer is copied down the process pipeline at the end of each update period
and, so, is available to by queried by other processes. The CUR statistics buffer is the
current working area and contains the statistics accumulated so far from previous stages
current frame; the contents of the CUR buffer is very dependent on the multiprocess
configuration (but is almost always empty in the application process, so queries should
access the PREV buffer). Statistics that are added to the CUR buffer by copying,
accumulation, or immediate-mode collection (such as with pfStatsCountGSet() and
pfFStatsCountNode()) are propagated down the process pipeline and then back up to
the application process to be included in the PREV buffer.
In a libpf application, most statistics collection is completely automatic. The
application must simply enable the desired classes of statistics with pfStatsClass()
and/or pfStatsClassMode().
The OpenGL Performer processes are responsible for actually opening the pfFrameStats
structure in which to accumulate the enabled statistics classes as well as for managing
any statistics hardware resources. All types of libpf statistics can be accumulated
without ever making specific calls to open a structure for accumulation or enabling
statistics hardware.
When using only libpr statistics, however, one must explicitly open a pfStats structure
for statistics accumulation by calling pfOpenStats().
Hardware statistics resources must also be managed by an application using only libpr
statistics. Statistics function calls that have “HW” in their names, such as
pfEnableStatsHw() and pfStatsHwAttr(), directly access system hardware (such as
graphics hardware and CPU); be careful to make such calls only from the relevant
process. pfEnableStatsHw() expects PFSTATSHW_EN* bitmask tokens. Statistics classes
which have corresponding statistics hardware have a PFSTATSHW_ prefix in their token
names.
In a libpf application, OpenGL Performer takes care of enabling the correct hardware
modes that correspond to enabled classes of statistics. For more information about
specific statistics classes, see the pfFrameStats(3pf) and pfStats(3pf) man pages.

007-1680-080

649

20: Statistics

Reducing the Cost of Statistics
Collecting and displaying statistics can have a big impact on performance. This section
describes ways to reduce that impact.
Enabling Only Statistics of Interest

Each channel has its Process Frame Times (PFTIMES) statistics class enabled by default.
This class maintains a short history of process frame times and averages the frame times
over the default update period of two seconds.
To minimize unnecessary overhead, turn off statistics on channels when you are not
using them. To turn off all statistics for a channel, call pfFStatsClass() in the application
process with the statistics structure of the given channel:
pfFStatsClass(pfGetChanFStats(chan), PFSTATS_ALL,
PFSTATS_OFF);

Each statistics class has default mode settings. The short history of process frame time is
used to draw the timing graph. By default, this history consists of four frames of each
OpenGL Performer process (app, cull, draw, intersections).
Maintaining this short history of statistics can be disabled by calling pfStatsClassMode()
with the token PFFSTATS_PFTIMES_HIST:
pfStatsClassMode(fstats, PFFSTATS_PFTIMES,
PFFSTATS_PFTIMES_HIST, PFSTATS_OFF);

This is useful if you are only interested in the average frame times of each task with
minimal overhead and you do not need to display the timing graph. However, for most
applications, the overhead incurred by keeping the timing history is not noticeable.
Controlling Update Rate

The update rate controls how often statistics are averaged and new results are made
available in the AVG buffer for display or query. Change the update rate by using the
following call:
pfFStatsAttr(fstats, {PFFSTATS_UPDATE_FRAMES,
PFFSTATS_UPDATE_SECS}, val);

650

007-1680-080

Collecting and Accessing Statistics in Your Application

When the update rate is nonzero, statistics are accumulated every frame. When the
update period is set to zero, no statistics accumulation or averaging is done and only
statistics in the PREV and CUR buffers are maintained.
When statistics are accumulated and averaged, the averaging happens only in the
application process, but accumulation is done in each OpenGL Performer process.

Statistics Output
Once you have collected some statistics, you need to be able to access and manipulate
them.
Printing

To print the contents of pfStats and pfFrameStats structures, use the general pfPrint()
routine. The verbosity-level parameter to pfPrint() sets the level of detail to use in
printing statistics. Statistics class-enable bitmasks can be used to select a subset of
statistics to print. For instance, to print only the enabled statistics, use the following:
pfPrint(stats, pfGetStatsClass(stats, PFSTATS_ALL),
PFPRINT_VB_INFO, 0);

When printing the contents of pfFrameStats structures, you can select which buffers are
to be printed: PREV, CUR, AVG, or CUM. The selected statistics from all selected buffers
are printed. The following call prints the currently enabled statistics from the previous
frame and from the averaged statistics buffer:
pfPrint(stats, PFFSTATS_BUF_PREV | PFFSTATS_BUF_AVG |
pfGetStatsClass(stats, PFSTATS_ALL),
PFPRINT_VB_INFO, 0);

Copying

You can copy entire pfStats and pfFrameStats structures with the general pfCopy()
command. pfCopy() copies all of the statistics data as well as information on mode
settings and which classes are enabled. The source and destination structures must be of
the same type. If both statistics structures are pfFrameStats structures, then all statistics
from all buffers are copied.

007-1680-080

651

20: Statistics

The pfCopyStats() and pfCopyFStats() routines copy only statistics data (not class
enables or mode settings) and accept a class enable bitmask to select statistics classes for
the copy, as shown in the following:
pfCopyStats(statsA, statsB, pfGetStatsClass(statsB,
PFSTATS_ALL));

For a pfFrameStats structure, a PFFSTATS_BUF_* token can be included in the stats
enable bitmask to select the buffer to be accessed. If no buffer is specified, the CUR buffer
is used. The following call copies the currently enabled classes of stats to the PREV
pfStats in fstats:
pfCopyFStats(fstats, stats, PFFSTATS_BUF_PREV |
pfGetStatsClass(stats, PFSTATS_ALL));

In this case, it is an error to select more than one statistics buffer; so, PFSTATS_ALL
cannot be used as the select. If you specify two pfFrameStats structures, the buffer select
is used for both structures; if you select multiple buffers, then each selected statistics class
from each selected buffer is copied. The pfCopyFStats() routine allows you to copy
between two different buffers of two pfFrameStats structures.
This routine takes explicit specification of PFFSTATS_BUF_* selects for source and
destination. Any PFFSTATS_BUF_* included with the class-enable bitmask is simply
ignored, making it safe to specify PFSTATS_ALL. This routine will not accept a pfStats
structure.
Querying

pfQueryStats() and pfMQueryStats() (and corresponding pfFrameStats versions) can be
used to get values from a pfStats or pfFrameStats structure and into an exposed structure.
These routines are useful when you want to use specific statistics for your own custom
load management or for benchmarking, and you can use them to implement your own
custom statistics utility routines. pfQueryStats() and pfMQueryStats() both take a
pfStats (or pfFrameStats for pfQueryFStats() and pfMQueryFStats()) and return the
number of bytes written to the provided destination buffer. pfQueryStats() takes a token
that specifies a single query while pfMQueryStats() expects a token buffer for multiple
queries. If an error is encountered, both query routines immediately halt and return with
the total number of bytes successfully written.

652

007-1680-080

Collecting and Accessing Statistics in Your Application

There are specific tokens for querying individual values or entire classes of statistics. The
query tokens are of the forms PFSTATSVAL_* and PFFSTATSVAL_*, and the
corresponding exposed structure names are of the form pfStatsVal* and pfFStatsVal*.
Queries on pfFrameStats structures with PFFSTATSVAL_* tokens expect a
PFFSTATS_BUF_* select token to be ORed with the query select. It is an error to include
more than one pfFrameStats buffer select token. If no buffer select token is provided, the
CUR buffer will be queried. The statistics query tokens and structures are defined in
prstats.h and pfstats.h.

Customizing Displays
The standard statistics displays have several parameters hard-wired. For instance, you
cannot change the colors used in such displays. If you want to use different colors, you
will have to use your own display routines.

Setting Update Rate
To set the frequency at which statistics are automatically collected, use pfFStatsAttr().
See the pfFrameStats(3pf) man page for details. If you want to turn off cumulative
statistics collection (and, thus, running averages) entirely, set the update rate to zero.
(Note that doing this will change your statistics display; in particular, your actual frame
rate will be changed and other averages will not be displayed.)

The pfStats Data Structure
The pfStats data structure contains four statistics buffers: one for current statistics, one
for previous statistics, one for cumulative statistics, and one for averages.
If you are using libpf calls to have OpenGL Performer keep track of statistics for you,
you should always look at the previous-stats buffer; the current-stats buffer is kept in a
state of flux, and if you look at it you are likely to find meaningless numbers there.
If, on the other hand, you are using libpr and keeping track of your own statistics, the
current-stats buffer does contain accurate information.

007-1680-080

653

20: Statistics

Setting Statistics Class Enables and Modes
This section contains some examples of statistics calls.
•

Set all statistics class enables on a pfStats to their default values:
pfStatsClass(stats, PFSTATS_ALL, PFSTATS_DEFAULT);

•

Set all modes for the PFSTATS_GFX class on a pfFrameStats to their default values:
pfFStatsClassMode(fstats, PFSTATS_GFX, PFSTATS_ALL,
PFSTATS_DEFAULT);

Note that pfStatsClassMode() takes a class name as its class specifier (second
argument) and not a bitmask. However, you can use PFSTATS_CLASS to refer to
the modes of all classes.
•

Set all modes of all pfStats classes to their default values:
pfFStatsClassMode(fstats, PFSTATS_MODE, PFSTATS_ALL,
PFSTATS_DEFAULT);

For pfFrameStats classes, there is PFFSTATS_CLASS.
•

Set the entire class enable mask to all PFSTATS_ALL, effectively enabling all
statistics classes:
pfFStatsClass(fstats, PFFSTATS_ALL, PFSTATS_SET);

•

Force off all modes of the PFSTATS_GFX class of a pfStats:
pfStatsClassMode(stats, PFSTATS_GFX, PFSTATS_OFF,
PFSTATS_SET);

•

To track triangle strip lengths on a pfFrameStats, enable the graphics statistics class
mode:
pfFStatsClassMode(fstats, PFSTATS_GFX,
PFSTATS_GFX_TSTRIP_LENGTHS, PFSTATS_ON);

654

007-1680-080

Chapter 21

21. Performance Tuning and Debugging

This chapter provides some basic guidelines to follow when tuning your application for
optimal performance. It also describes how to use debugging tools like pixie, prof,
and ogldebug to debug and tune your applications on IRIX. It concludes with some
specific notes on tuning applications on systems with RealityEngine graphics.

Performance-Tuning Overview
This section contains some general performance-tuning principles. Some of these issues
are discussed in more detail later in this chapter.

007-1680-080

•

Remember that high performance does not come by accident. You must design your
programs with speed in mind for optimal results.

•

Tuning graphical applications, particularly OpenGL Performer applications,
requires a pipeline-based approach. You can think of the graphics pipeline as
comprising three separate stages; the pipeline runs only as fast as the slowest stage;
so, improving the performance of the pipeline requires improving the slowest
stage’s performance. The three stages are the following:
–

The host (or CPU) stage, in which routines are called and general processing is
done by the application. This stage can be thought of as a software pipeline,
sometimes called the rendering pipeline, itself comprising up to three
sub-stages—the application, cull, and draw stages—as discussed at length
throughout this guide.

–

The transformation stage, in which transformation matrices are applied to
objects to position them in space (this includes matrix operations, setting
graphics modes, transforming vertices, handling lighting, and clipping
polygons).

–

The fill stage, which includes clearing the screen and then drawing and filling
polygons (with operations such as Gouraud shading, z-buffering, and texture
mapping).

655

21: Performance Tuning and Debugging

•

You can estimate your expected frame rate based on the makeup of the scene to be
drawn and graphics speeds for the hardware you are using. Be sure to include fill
rates, polygon transform rates, and time for mode changes, matrix operations, and
screen clear in your calculations.

•

Measure both the performance of complex scenes in your database and of
individual objects and drawing primitive to verify your understanding of the
performance requirements.

•

Use the OpenGL Performer diagnostic statistics to evaluate how long each stage
takes and how much it does. See Chapter 20, “Statistics,” for more information.
These statistics are referred to frequently in this chapter.

•

Tuning an application is an incremental process. As you improve one stage’s
performance, bottlenecks in other stages may become more noticeable. Also, do not
be discouraged if you apply tuning techniques and find that your frame rate does
not change—frame rates only change by a field at a time (which is to say in
increments of 16.67 milliseconds for a 60 Hz display) while tuning may provide
speed increases of finer granularity than that. To see performance improvements
that do not actually increase frame rate, look at the times reported by OpenGL
Performer statistics on the cull and draw processes (see Chapter 20 for more
information).

How OpenGL Performer Helps Performance
OpenGL Performer uses many techniques to increase application performance. Knowing
about what OpenGL Performer is doing and how it affects the various pipeline stages
may help you write more efficient code. This section lists some of the things OpenGL
Performer can do for you.

Draw Stage and Graphics Pipeline Optimizations
During drawing, OpenGL Performer does the following:

656

•

Sets up an internal record of what routines and rendering methods are fastest for
the current graphics platform. This information can be queried in any process with
pfQueryFeature(). You can use this information at run time when setting state
properties on your pfGeoStates.

•

Has machine-dependent fast options for commands that are very
performance-sensitive. Use the _ON and _FAST mode settings whenever possible

007-1680-080

How OpenGL Performer Helps Performance

to get machine-dependent high-performance defaults. Some examples include the
following:

007-1680-080

•

pfAntialias(PFAA_ON)

•

pfTransparency(PFTR_ON)

•

pfDecal(PFDECAL_BASE_TEST)

•

pfTexFilter(tex, filt, PFTEX_FAST)

•

pfTevMode(tev, PFTEV_FAST)

•

Sets up default modes for drawing, multiprocessing, statistics, and other areas, that
are chosen to provide high scene quality and performance. Some rendering defaults
differ from GL defaults: backface elimination is enabled by default
(pfCullFace(PFCF_BACK)) and lighting materials use glColorMaterial() to
minimize the number of materials required in a database (pfMtlColorMode(mtl,
side, PFMTL_CMODE_AD)).

•

Uses a large number of specialized routines for rendering different kinds of objects
extremely quickly. There is a specialized drawing routine for each possible pfGeoSet
configuration (each choice of primitive type, attribute bindings, and index
selection). Each time you change from one pfGeoSet to another, one of these
specialized routines is called. However, this happens even if the new pfGeoSet has
the same configuration as the old one, so larger pfGeoSets are more efficient than
smaller ones—the function-call overhead for drawing many small pfGeoSets can
reduce performance. As a rule of thumb, a pfGeoSet should contain at least
4 triangles, preferably between 8 and 64. If the pfGeoSet is too large, it can reduce
the efficiency of other parts of the process pipeline.

•

Caches state changes, because applying state changes is costly in the draw stage.
OpenGL Performer accumulates state changes until you call one of the pfApply*()
functions, at which point it applies all the changes at once. Note that this differs
from the graphics libraries, in which state changes are immediate. If you have
several state changes to make in OpenGL Performer, set up all the changes before
applying them, rather than using the one-change-at-a-time approach (with each
change followed by an apply operation) that you might expect if you are used to
graphics library programming.

•

Evaluates state changes lazily—that is, it avoids making any redundant changes.
When you apply a state change, OpenGL Performer compares the new graphics
state to the previous one to see if they are different. If they are, it checks whether the
new state sets any modes. If it does, OpenGL Performer checks each mode being set
to see whether it is different from the previous setting. To take advantage of this

657

21: Performance Tuning and Debugging

feature, share pfGeoStates and inherit states from the global state wherever
possible. Set all the settings you can at the global level and let other nodes inherit
those settings, rather than setting each node’s attributes redundantly. To do this
within a database, you can set up pfGeoStates with your desired global state and
apply them to the pfChannel or pfScene with pfChanGState() or pfSceneGState().
You can do this through the database loader utilities in libpfdu trivially for a
given scene with pfdMakeSharedScene() or have more control over the process
with pfdDefaultGState(), pfdMakeSceneGState(), and pfdOptimizeGStateList().
•

Provides an optimized immediate mode rendering display list data type,
pfDispList, in libpr. The pfDispList type reduces host overhead in the drawing
process and requires much less memory than a graphics library display list. libpf
uses pfDispLists to pass results from the cull process to the draw process when the
PFMP_CULL_DL_DRAW mode is turned on as part of the multiprocessing model.
For more information about display lists, see “Display Lists” in Chapter 9; for more
information about multiprocessing, see “Successful Multiprocessing with OpenGL
Performer” in Chapter 5.

Cull and Intersection Optimizations
To help optimize culling and intersection, OpenGL Performer does the following:

658

•

Provides pfFlatten() to resolve instancing (via cloning) and static matrix
transformations (by pre-transforming the cloned geometry). It can be especially
important to flatten small pfGeoSets; otherwise, matrix transformations must be
applied to each small pfGeoSet at significant computational expense. Note that
flattening resolves only static coordinate systems, not dynamic ones, but that, where
desired, pfDCS nodes can be converted to pfSCS nodes automatically using the
OpenGL Performer utility function pfdFreezeTransforms(), which allows for
subsequent flattening. Using pfFlatten(), of course, increases performance at the
cost of greater memory use. Further, the function pfdCleanTree() can be used to
remove needless nodes: identity matrix pfSCS nodes, single child pfGroup nodes,
and the like.

•

Uses bounding spheres around nodes for fast culling—the intersection test for
spheres is much faster than that for bounding boxes. If a node’s bounding sphere
does not intersect the viewing frustum, there is no need to descend further into that
node. There are bounding boxes around pfGeoSets; the intersection test is more
expensive but provides greater accuracy at that level.

007-1680-080

How OpenGL Performer Helps Performance

•

Provides the pfPartition node type to partition geometry for fast intersection
testing. Use a pfPartition node as the parent for anything that needs intersection
testing.

•

Provides level-of-detail (LOD) capabilities in order to draw simpler (and thus
cheaper) versions of objects when they are too far away for the user to discern small
details.

•

Allows intersection performance enhancement by precomputation of polygon
plane equations within pfGeoSets. This pre-computation is in the form of a traversal
that is nearly always appropriate—only in cases of non-intersectable or dynamically
changing geometry might these cached plane equations be disadvantageous. This
optimization is performed by pfuCollideSetup() using the PFTRAV_IS_CACHE bit
mask value.

•

Sorts pfGeoSets by graphics state in the cull process, in order to minimize state
changes and flatten matrix transformations when libpf creates display lists to
send to the draw process (as occurs in the PFMP_CULL_DL_DRAW
multiprocessing mode). This procedure takes extra time in the cull stage but can
greatly improve performance when rendering a complex scene that uses many
pfGeoStates. The sorting is enabled by default; it can be turned off and on by calling
the function pfChanTravMode(chan, PFTRAV_CULL, mode) and including or
excluding the PFCULL_SORT token. See “pfChannel Traversal Modes” in Chapter 4
and “Sorting the Scene” in Chapter 4 for more information on sorting.

Application Optimizations
During the application stage, OpenGL Performer does the following:

007-1680-080

•

Divides the application process into two parts: the latency-critical section (which
includes everything between pfSync() and pfFrame()), where last-minute
latency-critical operations are performed before the cull of the current frame can
start; and the noncritical portion, after pfFrame() and before pfSync(). The critical
section is displayed in the channel statistics graph drawn with pfDrawChanStats().

•

Provides an efficient mechanism to automatically propagate database changes
down the process pipeline and provides pfPassChanData() for passing custom
application data down the process pipeline.

•

Minimizes overhead in copying database changes to the cull process by
accumulating unique changes and updating the cull once inside pfFrame(). This
updated period is displayed in the MP statistics of the channel statistics graph
drawn with pfDrawChanStats().

659

21: Performance Tuning and Debugging

•

Provides a mechanism for performing database intersections in a forked process:
pass the PFMP_FORK_ISECT flag to pfMultiprocess() and declare an intersection
callback with pfIsectFunc().

•

Provides a mechanism for performing database loading and editing operations in a
forked process, such as the DBASE process: pass the PFMP_FORK_DBASE flag to
pfMultiprocess() and declare an intersection callback with pfDBaseFunc().

Specific Guidelines for Optimizing Performance
While OpenGL Performer provides inherent performance optimization, there are
specific techniques you can use to increase performance even more. This section contains
some guidelines and advice pertaining to database organization, code structure and
style, managing system resources, and rules for using OpenGL Performer.

Graphics Pipeline Tuning Tips
Tuning the graphics pipeline requires identifying and removing bottlenecks in the
pipeline. You can remove a bottleneck either by minimizing the amount of work being
done in the stage that has the bottleneck or, in some cases, by reorganizing your
rendering to more effectively distribute the workload over the pipeline stages. This
section contains specific tips for finding and removing bottlenecks in each stage of the
graphics pipeline.
Host Bottlenecks

Here are some ways to minimize the time spent in the host stage of the graphics pipeline:
•

660

Function calls, loops, and other programming constructs require a certain amount
of overhead. To make such constructs cost-effective, make sure they do as much
work as possible with each invocation. For instance, drawing a pfGeoSet of triangle
strips involves a nested loop, iterating on strips within the set and triangles within
each strip; it therefore makes sense to have several triangles in each strip and
several strips in each set. If you put only two triangles in a pfGeoSet, you will spend
all that loop overhead on drawing those two triangles, when you could be drawing
many more with little extra cost. The channel statistics can display (as part of the
graphics statistics) a histogram showing the percentage of your database that is
drawn in triangle strips of certain lengths.

007-1680-080

Specific Guidelines for Optimizing Performance

•

Only bind vertex attributes that are actually in use. For example, if you bind
per-vertex colors on a set of flat-shaded quads, the software will waste work by
sending those colors to the graphics pipeline, which will ignore them. Similarly, it is
pointless to bind normals on an unlit pfGeoSet.

•

Nonindexed drawing has less host overhead than indexed drawing because
indexed drawing requires an extra memory reference to get the vertex data to the
graphics pipeline. This is most significant for primitives that are easily host-limited,
such as independent polygons or very short triangle strips. However, indexed
drawing can be very helpful in reducing the memory requirements of a very large
database.

•

Enable state sorting for pfChannels (this is the default). By sorting, the CPU does
not need to examine as many pfGeoStates. The graphics channel statistics can be
used to report the pfGeoSet-to-pfGeoState drawing ratio.

Transform Bottlenecks

A transform bottleneck can arise from expensive vertex operations, from a scene that is
typically drawn with many very tiny polygons, from a scene modeled with inefficient
primitive types, or from excessive mode or transform operations. Here are some tips on
reducing such bottlenecks:

007-1680-080

•

Connected primitives will have better vertex rates than independent primitives,
and quadrilaterals are typically much more efficient in vertex operations than
independent triangles are.

•

The expensive vertex operations are generally lighting calculations. The fastest
lighting configuration is always one infinite light. Multiple lights, local viewing
models, and local lights have (in that order) dramatically increasing cost. Two-sided
lighting also incurs some additional transform cost. On some graphics platforms,
texturing and fog can add further significant cost to per-vertex transformation. The
channel graphics statistics will tell you what kinds of lights and light models are
being used in the scene.

•

Matrix transforms are also fairly expensive and definitely more costly than one or
two explicit scale, translate, or rotate operations. When possible, flatten matrix
operations into the database with pfFlatten().

•

The most frequent causes of mode changes are glShadeModel(), textures, and
object materials. The speed of these changes depends on the graphics hardware;
however, material changes do tend to be expensive. Sharing materials among
different objects can be increased with the use of pfMtlColorMode(), which is
PFMTL_CMODE_AD by default. However, on some older graphics platforms (such

661

21: Performance Tuning and Debugging

as the Elan, Extreme, and VGX), the use of pfMtlColorMode() (which actually calls
the function glColorMaterial()) has some associated per-vertex cost and should be
used with some caution.
•

If your cull stage is not a bottleneck, make sure your pfChannels sort the scene by
graphics state. Even if you are running in single process mode, the extra time taken
to sort the database is often more than offset by the savings in rendering time. See
“Sorting the Scene” in Chapter 4 for more details on how to configure sorting.

Fill Bottlenecks

Here are some methods of dealing with fill-stage bottlenecks:
•

One technique to hide the cost of expensive fill operations is to fill the pipeline from
the back forward so that no part is sitting idle waiting for an upstream stage to
finish. The last stage of the pipeline is the fill stage; so, by drawing backgrounds or
clearing the screen using pfClearChan() first, before pfDraw(), you can keep the fill
stage busy. In addition, if you have a couple of large objects that reliably occlude
much of the scene, drawing them very early on can both fill up the back-end stage
and also reduce future fill work, because the occluded parts of the scene will fail a
z-buffer test and will not have to write z values to the z-buffer or go on to more
complex fill operations.

•

Use the pfStats fill statistics (available for display through the channel statistics) to
visualize your depth complexity and get a count of how many pixels are being
computed each frame.

•

Be aware of the cost of any fill operations you use and their relative cost on the
relevant graphics hardware configuration. Quick experiments you can do to test for
a fill limitation include:
–

Rendering the scene into a smaller window, assuming that doing so will not
otherwise affect the scene drawn (a non-zero pfChannel LOD scale will cause a
change in object LODs when you shrink the window).

–

Using pfOverride() to force the disabling of an expensive fill mode.

If either of these tests causes a significant reduction in the time to draw the scene,
then a significant fill bottleneck probably exists.
Note: Some features may invoke expensive modes that need to be used with caution
on some hardware platforms. pfAntialias() and pfTransparency() enable blending if
multisampling is not available. Globally enabling these functions on machines
without multisampling can produce significant performance degradation due to the

662

007-1680-080

Specific Guidelines for Optimizing Performance

use of blending for an entire scene. Blending of even opaque objects incurs its full
cost. pfDecal() may invoke stenciling (particularly if you have requested the decal
mode PFDECAL_BASE_HIGH_QUALITY or if there is no faster alternative on the
current hardware platform), which can cause performance degradations on some
systems. pfFeature() can be used to verify the availability and speed of these features
on the current graphics platform.
•

The cost of specific fill operations can vary greatly depending on the graphics
hardware configuration. As a rule of thumb, flat shading is much faster than
Gouraud shading because it reduces both fill work and the host overhead of
specifying per-vertex colors. Z-buffering is typically next in cost, and then stencil
and blending. On a RealityEngine, the use of multisampling can add to the cost of
some of these operations, specifically z-buffering and stenciling. See
“Multisampling” on page 680 for more information. Texturing is considered free on
RealityEngine and Impact systems but is relatively expensive on a VGX and is
actually done in the host stage on lower-end graphics platforms, such as Extreme
and XZ. Some of the low-end graphics platforms also implement z-buffering on the
host.

•

You may not be able to achieve benchmark-level performance in all cases for all
features. For instance, if you frequently change modes and you use short triangle
strips, you get much less than the peak triangle mesh performance quoted for the
machine. Fill rates are sensitive to both modes, mode changes, and polygon sizes.
As a general rule of thumb, assume that fill rates average around 70% of peak on
general scenes to account for polygon size and position as well as for pipeline
efficiency.

Process Pipeline Tuning Tips
These simple tips will help you optimize your OpenGL Performer process pipeline:

007-1680-080

•

Use pfMultiprocess() to set the appropriate process model for the current machine.

•

You usually should not specify more processes with pfMultiprocess() than there are
CPUs on the system. The default multiprocess mode (PFMP_DEFAULT) attempts
an optimal configuration for the number of unrestricted CPUs. However, if there are
fewer processors than significant tasks (consider APP, CULL, DRAW, ISECT, and
DBASE), you will want experiment with the different two-process models to find
the one that will yield the best overall frame rate. Use of pfDrawChanStats(),
described in Chapter 20, “Statistics,” will greatly help with this task.

663

21: Performance Tuning and Debugging

664

•

Put only latency-critical tasks between the pfSync() and pfFrame() calls. For
example, put latency-critical updates, like changes to the viewpoint, after pfSync()
but before pfFrame(). Put time-consuming tasks, such as intersection tests and
system dynamics, after pfFrame().

•

You will also want to refer to the IRIX REACT documentation for setting up a
real-time system.

•

For maximum performance, use the OpenGL Performer utilities in libpfutil for
setting non-degrading priorities and isolating CPUs (pfuPrioritizeProcs(),
pfuLockDownProc(), pfuLockDownApp(), pfuLockDownCull(), and
pfuLockDownDraw()). These facilities require that the application runs with root
permissions. The source code for these utilities is in
/usr/share/Performer/src/lib/libpfutil/lockcpu.c for IRIX and
Linux and in %PFROOT%/Src/lib/libpfutil/lockcpu.c for Microsoft
Windows. For an example of their use, see the sample source code in
/usr/share/Performer/src/pguide/libpf/C/bench.c for IRIX and Linux
and in %PFROOT%/Src/pguide/libpf/C/bench.c for Microsoft Windows.
For more information about priority scheduling and real-time programming on
IRIX systems, see the chapter of the IRIX System Programming Guide entitled “Using
Real-Time Programming Features” and the IRIX REACT technical report.

•

Make sure you are not generating any floating-point exceptions. Floating-point
exceptions can cause an individual computation to incur tens of times its normal
cost. Furthermore, a single floating point exception can lead to many further
exceptions in computations that use the exceptional result and can even propagate
performance degradation down the process pipeline. OpenGL Performer will detect
and tell you about the existence of floating point exceptions if your pfNotifyLevel()
is set to PFNFY_INFO or PFNFY_DEBUG. You can then run your application in
dbx (on IRIX) or gdb (on Linux) and your application will stop when it encounters
an exception, enabling you to trace the cause.

•

Make sure the main pipeline (APP, CULL, and DRAW processes) do not make
per-frame memory allocations or deallocations (asynchronous processes like
DBASE can do per-frame allocations). On IRIX, you can use the debugging library’s
tracing feature of libdmalloc run-time malloc to verify that no memory
allocation routines are being called. On Linux, use the Electric Fence (libefence).
See the “Memory Corruption and Leaks” section later in this chapter for more about
libdmalloc.

•

Minimize the amount of channel data allocated by pfAllocChanData() and call
pfPassChanData() only when necessary to reduce the overhead of copying the
data. Copying pointers instead of data is often sufficient.

007-1680-080

Specific Guidelines for Optimizing Performance

Cull Process Tips

Here are a couple of suggestions for tuning the cull process:
•

The default channel culling mode enables all types of culling. If your cull process is
your most expensive task, you may want to consider doing less culling operations.
When doing database culling, always use view-frustum culling (PFCULL_VIEW)
and usually use graphics library mode database sorting (PFCULL_SORT) and
pfGeoSet culling (PFCULL_GSET) as well:
pfChanTravMode(chan, root,
PFCULL_VIEW | PFCULL_GSET | PFCULL_SORT);

A cull-limited application might realize a quick gain from turning off pfGeoSet
culling. If you think your database has few textures and materials, you might turn
off sorting. However, if possible it would be better to try improving cull
performance by improving database construction. “Efficient Intersection and
Traversals” on page 667 discusses optimizing cull traversals in more detail.
•

Look at the channel-culling statistics for the following:
–

A large amount of the database being traversed by the culling process and
being trivially rejected as not being in the viewing frustum. This can be
improved with better spatial organization of the database.

–

A large number of database nodes being non-trivially culled. This can be
improved with better spatial organization and breakup of large pfGeodes and
pfGeoSets.

–

A surprising number of LODs in their fade state (the fade computations can be
expensive, particularly if channel stress management has been enabled).

•

Balance the database hierarchy with the scene complexity: the depth of the
hierarchy, the number of pfGeoStates, and the depth of culling. See “Balancing Cull
and Draw Processing with Database Hierarchy” on page 669 for details.

•

pfNodes that have significant evaluation in the cull stage include pfBillboards,
pfLightPoints, pfLightSources, and pfLODs.

Draw Process Tips

Here are some suggestions specific to the draw process:
•

007-1680-080

Minimize host work done in the draw process before the call to pfDraw(). Time
spent before the call to pfDraw() is time that the graphics pipeline is idle. Any

665

21: Performance Tuning and Debugging

graphics library (or X) input processing or mode changes should be done after
pfDraw() to take effect in the following frame.

666

•

Use only one pfPipe per hardware graphics pipeline and preferably one
pfPipeWindow per pfPipe. Use multiple channels within that pfPipeWindow to
manage multiple views or scenes. It is fairly expensive to simultaneously render to
multiple graphics windows on a single hardware graphics pipeline and is not
advisable for a real-time application.

•

Pre-define and pre-load all of the textures in the database into hardware texture
memory by using a pfApplyTex() function on each pfTexture. You can do this
texture-applying in the pfConfigStage() draw callback or (for multipipe
applications to allow parallelism) the pfConfigPWin() callback. This approach
avoids the huge performance impact that results when textures are encountered for
the first time while drawing the database and must then be downloaded to texture
memory. Utilities are provided in libpfutil to apply textures appropriately; see
the pfuDownloadTexList() routine in the distributed source code file
/usr/share/Performer/src/lib/libpfutil/tex.c for IRIX and Linux and
in file %PFROOT%/Src/lib/libpfutil/tex.c on Microsoft Windows. The
Perfly application demonstrates this. For IRIX and Linux, see the Perfly source file
generic.c in either the C language
(/usr/share/Performer/src/sample/C/common) or C++ language
(/usr/share/Performer/src/sample/C++/common) versions of Perfly.For
Microsoft Windows, see the Perfly source file generic.c in either the C language
(%PFROOT%/Src/sample/C/common) or C++ language
(%PFROOT%/Src/sample/C++/common) versions of Perfly.

•

Minimize the use of pfSCSs and pfDCSs and nodes with draw callbacks in the
database since aggressive state sorting is kept local to subtrees under these nodes.

•

Do not do any graphics library input handling in the draw process. Instead, use
X input handling in an asynchronous process. OpenGL Performer provides utilities
for asynchronous input handling in libpfutil with source code provided in
/usr/share/Performer/src/lib/libpfutil/input.c for IRIX and Linux
and in %PFROOT%/Src/lib/libpfutil/input.c for Microsoft Windows. For a
demonstration of asynchronous X input handling, see provided sample
applications, such as perfly, and also the distributed sample programs
/usr/share/Performer/src/pguide/libpf/C/motif.c and
/usr/share/Performer/src/pguide/libpfui/C/motifxformer.c for
IRIX and Linux and the programs
%PFROOT%/Src/lib/pguide/libpf/C/motif.c and
%PFROOT%/Src/lib/pguide/libpfui/C/motifxformer.c for Microsoft
Windows.

007-1680-080

Specific Guidelines for Optimizing Performance

Efficient Intersection and Traversals

Here are some tips on optimizing intersections and traversals:
•

Use pfPartition nodes on pieces of the database that will be handed to intersection
traversal. These nodes impose spatial partitioning on the subgraph beneath them,
which can dramatically improve the performance of intersection traversals.
Note: Subgraphs under pfDCS, pfLOD, pfSwitch, and pfSequence nodes are not
partitioned; so, intersection traversals of these subgraphs will not be affected.

•

Use intersection caching. For static objects, enable intersection caching at
initialization—first call pfNodeTravMask(), specifying intersection traversal
(PFTRAV_ISECT), and then include PFTRAV_IS_CACHE in the mode for
intersections. You can turn this mode on and off for dynamic objects as appropriate.

•

Use intersection masks on nodes to eliminate large sections of the database when
doing intersection tests. Note that intersections are sproc()-safe in the current
version of OpenGL Performer; you can check intersections in more than one
process.

•

Bundle segments for intersections with bounding cylinders. You can pass as many
as 32 segments to each intersection request. If the request contains more than a few
segments and if the segments are close together, the traversal will run faster if you
place a bounding cylinder around the segments using pfCylAroundSegs() and pass
that bounding cylinder to pfNodeIsectSegs(). The intersection traversal will use the
cylinder rather than each segment when testing the segments against the bounding
volumes of nodes and pfGeoSets.
Note: Bundling segments pertains only to IRIX and Linux.

Database Concerns
Optimizing your databases can provide large performance increases.
libpr Databases
The following tips will help you achieve peak performance when using libpr:
•

007-1680-080

Minimize the number of pfGeoStates by sharing as much as possible.

667

21: Performance Tuning and Debugging

•

Initialize each mode in the global state to match the majority of the database in
order to set as little local state for individual pfGeoStates as possible.

•

Use triangle strips wherever possible; they produce the largest number of polygons
from a given number of vertices; so, they use the least memory and are drawn the
fastest of the primitive types.

•

Use the simplest possible attribute bindings and use flat-shaded primitives
wherever possible. If you are not going to need an object’s attributes, do not bind
them—anything you bind will have to be sent to the pipeline with the object.

•

Flat-shaded primitives and simple attribute bindings reduce the transformation and
lighting requirements for the polygon. Note that the flat-shaded triangle-strip
primitive renders faster than a regular triangle strip, but you have to change the
index by two to get the colors right (that is, you need to ignore the first two vertices
when coloring). See “Attributes” in Chapter 8 for more information.

•

Use nonindexed drawing wherever possible, especially for independent polygon
primitives and short triangle strips.

•

When building the database, avoid fragmentation in memory of data to be
rendered. Minimize the number of separate data and index arrays. Keep the data
and index arrays for pfGeoSets contiguous and try to keep separate pfGeoSets
contiguous to avoid small, fragmented pfMalloc() memory allocations.

•

The ideal size of a pfGeoSet (and of each triangle strip within the pfGeoSet)
depends a great deal on the specific CPU and system architecture involved; you
may have to do benchmarks to find out what is best for your machine. For a general
rule of thumb, use at least 4 triangles per strip on any machine, and 8 on most. Use
5 to 10 strips in each pfGeoSet, or a total of 24 to 100 triangles per pfGeoSet.

libpf Databases
When you are using libpf, the following tips can improve the performance of database
tasks:

668

•

Use pfFlatten(), especially when a pfScene contains many small instanced objects
and billboards. Use pfdCleanTree() and (if application considerations permit)
pfdFreezeTransforms() to minimize the cull traversal processing time and
maximize state sorting scope.

•

Initialize each mode in the scene pfGeoState to match the majority of the database in
order to set as little local state for individual pfGeoStates as possible. The utility
function pfdMakeSharedScene() provides an easy to use mechanism for this task.

007-1680-080

Specific Guidelines for Optimizing Performance

•

Minimize the number of very small pfGeoSets (that is, those containing four or
fewer total triangles). Each tiny pfGeoSet means another bounding box to test
against if you are culling down to the pfGeoSet level (that is, when PFCULL_GSET
is set with pfChanTravMode()) as well as another item to sort during culling. (If
your pfGeoSets are large, on the other hand, you should definitely cull down to the
pfGeoSet level.)

•

Be sparing in the use of pfLayers. Layers imply that pixels are being filled with
geometry that is not visible. If fill performance is a concern, this should be
minimized in the modeling process by cutting layers into their bases when possible.
However, this will produce more polygons which require more transform and host
processing; so, it should only be done if it will not greatly increase database size.

•

Make the hierarchy of the database spatially coherent so that culling will be more
accurate and geometry that is outside the viewing frustum will not be drawn. (See
Figure 4-3 for an example of a spatially organized database.)

Balancing Cull and Draw Processing with Database Hierarchy

Construct your database to minimize the draw-process time spent traversing and
rendering the culled part of the database without the cull-process time becoming the
limiting performance factor. This process involves making tradeoffs as a simpler cull
means a less efficient draw stage. This section describes these tradeoffs and some good
rules to follow to give you a good start.
If the cull and draw processes are performed in parallel, the goal is to minimize the larger
of the culling and drawing times. In this case, an application can spend approximately
the same amount of time on each task. However, if both culling and drawing are
performed in the same process, the goal is to optimize the sum of these two times, and
both processes must be streamlined to minimize the total frame time. Important
parameters in this optimization include the number of pfGeoSets, the average branching
factor of the database hierarchy, and the enabled channel culling traversal modes. The
pfDrawChanStats() function (see Chapter 20, “Statistics”) can easily provide diagnostic
information to aid in this tuning.
The average number of immediate children per node can directly affect the culling
process. If most nodes have a high number of children, the bounding spheres are more
likely to intersect the viewing frustum and all those nodes will have to be tested for
visibility. At the other extreme, a very low number of children per node will mean that
each bounding sphere test can only eliminate a small part of the database and so many
nodes may still have to be traversed. A good place to start is with a quad-tree type
organization where each node has about four children and the bounding geometry of

007-1680-080

669

21: Performance Tuning and Debugging

sibling nodes are adjacent but have minimal intersection. In the ideal case, projected to a
two-dimensional plane on the ground, the spatial extent of each node and its parents
would form a hierarchy of boxes.
The transition from pfGeodes to pfGeoSets is an important point in the database
structure. If there are many very small pfGeoSets within a single pfGeode, not culling
down to pfGeoSets can actually improve overall frame time because the cost of drawing
the tiny pfGeoSets may be small relative to the time spent culling them. Adding more
pfGeodes to break up the pfGeoSets can help by providing a slightly more accurate cull
at less cost than considering each pfGeoSet individually. In addition, pfGeodes are culled
by their bounding spheres, which is faster than the culling of pfGeoSets, which are culled
by their bounding boxes.
The size (both spatial extend and number of triangles) can also directly impact culling
and drawing performance. If pfGeoSets are relatively large, there will be fewer to cull so
pfGeoSet culling can probably be afforded. pfGeoSets with more triangles will draw
faster. However, pfGeoSets with larger spatial extent are more likely to have geometry
being drawn outside of the viewing frustum; this wastes time in the graphics stage.
Breaking up some of the large pfGeoSets can improve graphics performance by allowing
a more accurate cull.
With some added cost to the culling task, the use of level-of-detail nodes (pfLODs) can
make a tremendous difference in graphics performance and image quality. LODs allow
objects to be automatically drawn with a simplified version when they are in a state that
yields little contribution to the scene (such as being far from the eyepoint). This allows
you to have many more objects in your scene than if you always were drawing all objects
at full complexity. However, you do not want the cull to be testing all LODs of an object
every frame when only one will be used. Again, proper use of hierarchy can help.
pfLODs (non-fading) can be inserted into the hierarchy with actual object pfLODs
grouped beneath them. If the parent LOD is out of range for the current viewpoint, the
child LODs will never be tested. The pfLODs of each object can be placed together under
a pfGroup so that no LOD tests for the object will be done if the object is outside of the
viewing frustum.
Calling pfFlatten(), pfdFreezeTransforms(), or pfdCleanTree() to remove extraneous
nodes can often help culling performance. Use pfFlatten() to de-instance and apply
pfSCS node transformations to leaf geometry—resulting in less work during the cull
traversal. This allows both better database sorting for the draw and also better caching
of matrix and bounding information, which can speed up culling. When these scene
graph modifications are not acceptable, you may reduce cull time by turning off culling

670

007-1680-080

Specific Guidelines for Optimizing Performance

of pfGeoSets but this will directly impact rendering performance by forcing the
rendering of geometry that is outside the viewing frustum.
Tip: Making the scene into a graphics library object in the draw callback can show the
result of the cull; this can give a visual check of what is actually being sent to the graphics
subsystem. Check for objects that are far from the viewing frustum; this can indicate that
the pfGeodes or pfGeoSets need to be broken up. Additionally, the rendering time of the
GL object should be compared to the pfDraw() rendering time to see if the pfGeoSets
have enough triangles in them to not incur host overhead. Alternately, the view frustum
can be made larger than that used in the cull to allow simple cull volume visualization
during real-time simulation. The OpenGL Performer sample program perfly supports
this option. Press the z key while in Perfly to enable cull volume visualization and inspect
the resulting images for excessive off-screen geometric content. Such content is a clear
sign that the database could profitably be further subdivided into smaller components.

Graphics and Modeling Techniques to Improve Performances

On machines with fast-texture mapping, texture should be used to replace complex
geometry. Complex objects, such as trees, signs, and building fronts, can be effectively
and efficiently represented by textured images on single polygons in combination with
applying pfAlphaFunc() to remove parts of the polygon that do not appear in the image.
Using texture to reduce polygonal complexity can often give both an improved picture
and improved performance. This is because of the following:
•

The image texture provides scene complexity, and the texture hardware handles
scaling of the image with MIPmap interpolation functions for minification (and, on
RealityEngine systems, Sharpen and DetailTexture functions for magnification).

•

Using just a few textured polygons rather than a complex model containing many
individual polygons reduces system load.

In order to represent a tree or other 3D object as a single textured polygon, OpenGL
Performer can rotate polygons to always face the eyepoint. An object of this type is
known as a billboard and is represented by a pfBillboard node. As the viewer moves
around the object, the textured polygon rotates so that the object appears
three-dimensional. For more information on billboards, see “pfBillboard Nodes” in
Chapter 3.
To determine if the current graphics platform has fast-texture mapping, look for a
PFQFTR_FAST return value from the following call:

007-1680-080

671

21: Performance Tuning and Debugging

pfFeature(PFQFTR_TEXTURE, &ret);

pfAlphaFunc() with a function PFAF_GEQUAL and a reference value greater than zero
can be used whenever transparency is used to remove pixels of low contribution and
avoid their expensive processing phase.

Special Coding Tips
For maximum performance, routines that make extensive use of the OpenGL Performer
linear algebra routines should use the macros defined in prmath.h to allow the compiler
to in-line these operations.
Use single- rather than double-precision arithmetic where possible and avoid the use of
short-integer data in arithmetic expressions. Append an ‘f’ to all floating point constants
used in arithmetic expressions.
BAD

In this example, values in both of the expressions involving the floating
point variable x are promoted to double precision when evaluated:
float x;
if (x < 1.0)
x = x + 2.0;

GOOD

In this example, both of the expressions involving the floating point
variable x remain in single-precision mode, because there is an ‘f’
appended to the floating point constants:
float x;
if (x < 1.0f)
x = x + 2.0f;

Performance Measurement Tools
Performance measurement tools can help you track the progress of your application by
gathering statistics on certain operations. OpenGL Performer provides run-time
profiling of the time spent in parts of the graphics pipeline for a given frame. The
pfDrawChanStats() function displays application, cull, and draw time in the form of a
graph drawn in a channel; see Chapter 20, “Statistics,” for more information on that and
related functions. There are advanced debugging and tuning tools available from SGI
that can be of great assistance.The WorkShop product in the CASEVision tools provides
a complete development environment for debugging and tuning of host tasks. The

672

007-1680-080

Performance Measurement Tools

Performance Co-Pilot helps you to tune host code in a real-time environment. There is
also the WindView product from WindRiver that works with IRIX REACT to do full
system profiling in a real-time environment. However, progress can be made with the
basic tools that are in the IRIX development environment: prof and pixie. On Linux,
use gprof and on Microsoft Windows use VTune (see
www.intel.com/software/products/vtune). The OpenGL debugging utility,
ogldebug, can also be used to aid in performance tuning. This section briefly discusses
getting started with these tools.
Note: See the graphics library manual available from SGI for complete instructions on
using these graphics tools. See the IRIX System Programming Guide to learn more about
pixie and prof.

Using pixie, prof, and gprof to Measure Performance
Note: This section does not pertain to Microsoft Windows systems.
You can use the IRIX performance analysis utilities pixie and prof to tune the
application process. For Linux, use gprof. Use pixie for basic-block counting and use
prof or gprof for program counter (PC) sampling. PC sampling gives run-time
estimation of where significant amounts of time are spent, whereas basic-block counting
will report the number of times a given instruction is executed.
To isolate statistics for the application process, even in single-process models, run the
application through pixie or prof in APP_CULL_DRAW mode to separate out the
process of interest. Both pixie and prof can generate statistics for an individual
process.
When using OpenGL Performer DSO libraries with prof you may want to provide the
-dso option to prof with the full pathname of the library of interest to have OpenGL
Performer routines included in the analysis. When using pixie, you will need to have
the .pixie versions of the DSO libraries in your LD_LIBRARY_PATH. Additionally,
you will need a .pixie version of the loader DSO for your database in your
LD_LIBRARY_PATH. You may have to pixie the loader DSO separately since pixie
will not find it automatically if your executable was not linked with it. When using prof
to do PC sampling, link with unshared libraries exclusively and use the –p option to ld.

007-1680-080

673

21: Performance Tuning and Debugging

Then set the environment variable PROFDIR to indicate the directory in which to put
profiling data files.
When profiling, run the program for a while so that the initialization will not be
significant in the profiling output. When running a program for profiling, run a set
number of frames and then use the automatic exit described below.

Using ogldebug to Observe Graphics Calls
You can use the graphics utility ogldebug to both debug and tune OpenGL Performer
applications. Note that ogldebug can handle multiprocessed programs.
Use ogldebug to do the following:
•

Show which graphics calls are being issued.

•

Look for frequent mode changes or unnecessary mode settings that can be caused if
your initialization of the global state does not match the majority of the database.

•

Look for unnecessary vertex bindings such as unneeded per-vertex colors or
normals for a flat-shaded object.

Follow these steps to examine one frame of the application in an ogldebug session:
1.

Start the profiler of your choice:
ogldebug your_prog_name prog_options

2. Turn off output and breakpoints from the control panel.
3. Under IRIX and Linux, set a breakpoint at glXSwapBuffers().Under Microsoft
Windows, set a breakpoint at wglSwapBuffers().
4. Click the “Continue” button and go to the frame of interest.
5. Turn on breakpoints.
Execution stops at glXSwapBuffers().
6. Turn on all trace output.
7. Click the “Continue” button.
Execution stops at the next glSwapBuffers(), outputting one full scene to
progname.pid.trace.
8. Quit and examine the output.

674

007-1680-080

Guidelines for Debugging

Note: Since OpenGL Performer avoids unnecessary mode settings, recording one frame
shows modes that are set during that frame, but it does not reflect modes that were set
previously. It is, therefore, best to have a program that can come up in the desired
location and with the desired modes, then grab the first two frames: one for initialization
and one for continued drawing.

Guidelines for Debugging
This section lists some general guidelines to keep in mind when debugging OpenGL
Performer applications.

Shared Memory
Because malloc() does not allocate memory until that memory is used, core dumps may
occur when arenas do not find enough disk space for paging. On IRIX systems, the kernel
can be configured to actually allocate space upon calling malloc(), but this change is
pervasive and has performance ramifications for fork() and exec(). Reconfiguring the
kernel is not recommended, so be aware that unexplained core dumps can result from
inadequate disk space.
Be sure to initialize pointers to shared memory and all other nonshared global values
before OpenGL Performer creates the additional processes in the call to pfConfig().
Values put in global variables initialized after pfConfig() will only be visible to the
process that set them.
For detailed information about other aspects of shared memory, see “Memory
Allocation” in Chapter 15.

Use the Simplest Process Model
When debugging an application that uses a multiprocess model, first use a single-process
model to verify the basic paths of execution in the program. You do not have to
restructure your code; simply select single-process operation by calling
pfMultiprocess(PFMP_APPCULLDRAW) to force all tasks to initiate execution
sequentially on a frame-by-frame basis.

007-1680-080

675

21: Performance Tuning and Debugging

If an application fails to run in multiprocess mode but runs smoothly in single-process
mode, you may have a problem with the initialization and use of data that is shared
among processes.
If you need to debug one of multiple processes, use the following command while the
process is running:
IRIX% dbx -p progname
LINUX% gdb progname pid

For Microsoft Windows, use the following commands:
set PATH=%PFROOT%\Lib\Debug;%PATH%
msdev programName programParameters

The preceding is how you debug using Microsoft Developer Studio 6.0.
This will show the related processes and allow you to choose a process to trace. The
application process will always be the process with the lowest process ID. In order after
that will be the (default) clock process, then the cull process, and then the draw.
Once the program works, experiment with the different multiprocess models to achieve
the best overall frame rate for a given machine. Do not specify more processes than
CPUs. Use pfDrawChanStats() to compare the frame timings of the different stages and
frame times for the different process models.

Avoid Floating-Point Exceptions
Arrange error handling for floating-point operations. To see floating-point errors, turn
debug messages on and enable floating-point traps. Set
pfNotifyLevel(PFNFY_DEBUG).
The goal is to have no NaN (Not a Number), INF (infinite value), or floating-point
exceptions resulting from numerical computations.

When the Debugger Will Not Give You a Stack Trace
If a NULL or invalid function pointer is called, the program dies with a segmentation
fault, bus error, or invalid instruction, and the debugger is often unable to give a stack
trace.

676

007-1680-080

Guidelines for Debugging

(dbx or gdb) where
>

0 () [< unknown >, 0x0]

When this happens on IRIX, you can usually still get a single level of trace by looking at
the return address register.
(dbx) $ra/i
[main:6, 0x400c18]

ra,28(sp)

Once you know this information, you may be able to set a breakpoint to stop just before
the bad function pointer and then get a full stack trace from there.
On Microsoft Windows, do the following:
1.

Ensure that the path to the debug libraries is set:
set PATH=%PFROOT%\Lib\Debug;%PATH%

2. Run the program from your normal development environment.
3. Enter the following command:
set _PF_DISPLAY_MODULES_LOADED=1

Tracing Members of OpenGL Performer Objects
Debuggers like dbx and gdb allow you to set a breakpoint or trace on a particular
variable or address in memory.
However, this feature does not work well on programs that use atomic shared memory
access functions like test_and_set() which are implemented on IRIX using the MIPS
instructions ll and sc. Calling such a function on an address that is on the same memory
page (typically 4096 bytes) as the address where a breakpoint is set results in the program
being killed with a SIGTRAP signal.
OpenGL Performer uses test_then_add() to implement pfMemory::ref() and
pfMemory::unref(); so, you almost always run into this problem if you try to trace a
member of an OpenGL Performer object (anything derived from pfMemory).
You can get around the problem by setting the following environment variable before
running the program.

007-1680-080

677

21: Performance Tuning and Debugging

Under IRIX and Linux:
% setenv _PF_OLD_REFCOUNTING

Under Microsoft Windows:
set _PF_OLD_REFCOUNTING=1

This tells OpenGL Performer to use an alternate (slower) implementation of shared
memory reference counting that avoids the ll and sc operations so that breakpoints on
variables can be used.

Memory Corruption and Leaks
A number of tools are available for debugging memory corruption errors and memory
leaks; no single one is ideal for all purposes. We will briefly describe and compare two
useful tools: purify and libdmalloc.

Purify
Purify is a product of PureAtria; see their Web site, http://www.pureatria.com for
ordering information.
Purify works by rewriting your program (and all dynamic shared libraries it uses),
intercepting malloc and associated functions, and inserting instruction sequences
before each load and store instruction that immediately catch invalid memory accesses
(that is, uninitialized memory reads, out-of-bounds memory reads or writes, freed
memory reads or writes, and multiple frees) and also keeps track of memory leaks.
When an error is encountered, a stack trace is given for the error as well as a stack trace
from when the memory in question was originally allocated. When used in conjunction
with a debugger like dbx or gdb , you can set a breakpoint to stop when Purify
encounters an error so that you can examine the program state.
Purify is immensely useful for tracking down memory problems. Its main drawbacks are
that it is very slow (to compile and run programs) and it currently does not know about
the functions amalloc(), afree(),and arealloc(), which are very important in
OpenGL Performer applications (pfMalloc and associated functions are implemented in
terms of them).

678

007-1680-080

Memory Corruption and Leaks

You can trick Purify into telling you about some shared arena memory access errors by
telling the run-time linker to use malloc in place of amalloc; however, this cannot
work if there are forked processes sharing the arena, so to use this trick you must run in
single-process (PFMP_APPCULLDRAW) mode. To do this, compile the following into a
DSO called pureamalloc.so:
malloc(int n) {return malloc(n);}
afree(void*p) {free(p);}
arealloc(void*p, int n) {return realloc(p, n);}

Then tell the run-time linker to point to it:
% setenv _RLDN32_LIST `pwd`/pureamalloc.so:DEFAULT

This assumes you are using the N32 ABI; see the rld man page for the equivalents for
the O32 and 64 ABIs.Then run perfly -m0 or any other program in
PFMP_APPCULLDRAW mode.
For more information on Purify, visit PureAtria’s web site:
http://www.pureatria.com.

libdmalloc (IRIX only)
libdmalloc is a library that was developed internally at SGI. It is officially
unsupported but you can get it through OpenGL Performer’s ftp site.
libdmalloc is implemented as a dynamic shared object (DSO) that you can link in to
your program at run-time. It intercepts all calls to malloc(), free(), and associated
functions and checks for memory corruption of the particular piece of memory being
accessed. It also attempts to purposely break programs that make bad assumptions: it
initializes newly malloced (or amalloced) memory with a fill pattern to break programs
that depend on it being 0’s and similarly fills freed memory with a fill pattern to break
programs that look at freed memory. Finally, the entire malloc arena and any shared
arenas are checked for corruption during exit() and execve(). Error messages are printed
to stderr whenever an error is detected (which is typically later than the error actually
occured, unlike Purify’s immediate detection). libdmalloc does not know about
UMRs (uninitialized memory reads) or stack variables, unlike Purify.
libdmalloc’s advantages are that it knows about amalloc/afree/arealloc, and it has
virtually no overhead; so, you can leave it on all the time (it will check for errors in every
program you run). If it uncovers a reproducible bug in an area that Purify knows about

007-1680-080

679

21: Performance Tuning and Debugging

too, then you can use Purify to trace the exact location of the problem— Purify is much
better at that than libdmalloc.
When using libdmalloc, you can easily toggle verbose tracing of all calls to
malloc()/free()/realloc() and amalloc()/afree()/arealloc() by sending the processes a
signal at run time— there should be none of these calls per-frame in the main pipeline of
a tuned OpenGL Performer application.
For more information, install libdmalloc from the OpenGL Performer ftp site and
read the files /usr/share/src/dmalloc/README and, for OpenGL
Performer-specific suggestions,
/usr/share/src/dmalloc/SOURCEME.dmalloc.performer.

Notes on Tuning for RealityEngine Graphics
This section contains some specific notes on performance tuning with RealityEngine
graphics.

Multisampling
Multisampling provides full-scene antialiasing with performance sufficient for a
real-time visual simulation application. However, it is not free and it adds to the cost of
some other fill operations. With RealityEngine graphics, most other modes are free until
you add multisampling— multisampling requires some fill operations to be performed
on every subpixel. This is most noticeable with z-buffering and stenciling operations but
also applies to glBlendFunc(). Texturing is an example of a fill operation that can be free
on a RealityEngine and is not affected by the use of multisampling.
The multisampling hardware reduces the cost of subpixel operations by optimizing for
pixels that are fully opaque. Pixels that have several primitives contributing to their
result are thus more expensive to evaluate and are called complex pixels. Scenes usually
end up having a very low ratio of complex pixels.
Multisampling offers an additional performance optimization that helps balance its cost:
a virtually free screen clear. Technically, it does not really clear the screen but rather
allows you to set the z values in the framebuffer to be undefined. Therefore, use of this
clear requires that every pixel on the screen be rendered every frame. This clear is

680

007-1680-080

Notes on Tuning for RealityEngine Graphics

invoked with a pfEarthSky using the PFES_TAG option to pfESkyMode(). Refer to the
pfEarthSky(3pf) man page for more detailed information.

Transparency
There are two ways of achieving transparency on a RealityEngine: blending and
screen-door transparency with multisampling.
Blended transparency, using the routine glBlendFunc(), can be used with or without
multisampling. Blending does not increase the number of complex pixels but is
expensive for complex pixels.
To reduce the number of pixels with very low alpha, one can use a pfAlphaFunc() that
ignores pixels of low alpha, such as alpha less than 3 or 4. This will slightly improve fill
performance and probably not have a noticeable effect on scene quality. Many scenes can
use values as high as 60 or 70 without suffering degradation in image quality. In fact, for
a scene with very little actual transparency, this can reduce the fuzzy edges on textures
that simulate geometry (such as trees and fences) that arise from MIP-mapping.
Screen-door transparency gives order-independent transparent effects and is used for
achieving the fade-LOD effect. It is a common misperception that screen-door
transparency on RealityEngine gives you n levels of transparency for n multisamples. In
fact, n samples gives you 4n levels of transparency, because RealityEngine uses 2-pixel
by 2-pixel dithering. However, screen-door transparency causes a dramatic increase in
the number of complex pixels in a scene, which can affect fill performance.

Texturing
Texturing is free on a RealityEngine if you use a 16-bit texel internal texture format. There
are 16-bit texel formats for each number of components. These formats are used by
default by OpenGL Performer but can be set on a pfTexture with pfTexFormat(). Using a
32-bit texel format will yield half the fill rate of the 16-bit texel formats.
Do not use huge ranges of texture coordinates on individual triangles. This can incur
both an image quality degradation and a severe performance degradation. Keep the
maximum texture coordinate range for a given component on a single triangle under the
following value:
1 << (13-log2(TexCSize))

007-1680-080

681

21: Performance Tuning and Debugging

where TexCSize is the size in the dimension of that component.
The use of Detail Texture and Sharpen can greatly improve image quality. Minimize the
number of different detail and sharpen splines (or just use the internal default splines).
Applying the same detail texture to many base textures can incur a noticeable cost when
base textures are changed. Detail textures are intended to be derived from a
high-resolution image that corresponds to that of the base texture.

Other Tips
Two final notes on RealityEngine performance:
•

Changing the width of antialiased lines and points is expensive.

•

pfMtlColorMode() (which calls the function glColorMaterial()) has a huge
performance benefit.

EventView—A Performance Analyzer
It is often very difficult to analyze a single missed frame by using the pfChannel statistics
displays. OpenGL Performer is instrumented to generate internal timing events and
write them into a file upon an application program request. You can analyze these
collected events at your own pace using a graphic analysis tool. EventView is such a tool.
Note: EventView was originally developed by Ran Yakir from BVR Systems.
Each event has a start time and an end time. Together they bind the execution of
important parts of an OpenGL Performer frame in the various OpenGL Performer
processes. The captured event durations are similar to the durations presented in
pfFrameStats. However, EventView events can be written to a file and analyzed offline
after the execution of an OpenGL Performer program completes. In addition, an
application program can design its private set of events and can merge them into the
same output file as the OpenGL Performer internal events.
The following subsections describe the use of EventView:
•

682

“Viewing Events—evanalyzer” on page 683

007-1680-080

EventView—A Performance Analyzer

•

“Controlling the Collection of OpenGL Performer Internal Events” on page 684

•

“Sample Use of EventView” on page 684

•

“Using EventView Tools” on page 688

•

“Understanding OpenGL Performer Internal Events” on page 695

Viewing Events—evanalyzer
EventView provides an analysis program called evanalyzer for generating a graphic
display of the captured events. Figure 21-1 shows a sample screen from the evanalyzer
program. Each row contains events of a single type. For each event, its start is marked by
a vertical line, its duration is marked by a horizontal line, and its end is marked by
another vertical line extending upward.

Figure 21-1

007-1680-080

The evanalyzer Main Display

683

21: Performance Tuning and Debugging

Controlling the Collection of OpenGL Performer Internal Events
Use the following functions in order to control EventView event collection:
Function

Description

pfInitializeEvents()

Initializes the internal event capturing mechanism of
OpenGL Performer. An application must call this function
before calling pfConfig() in order to initialize the
event-capturing data structures. All other event-capturing
functions should be called after pfInitializeEvents().

pfEventSampleOn()

Starts event capturing. Immediately after this function is
called, all timing events generated by OpenGL Performer are
logged for future saving into a file. OpenGL Performer will
log a limited number of events (currently 10,000 events). No
events will be logged after this limit is reached.

pfEventSampleOff()

Stops event capturing. No events will be logged after this call.

pfWriteEvents()

Writes the currently logged events into a disk filename. The
written events can be viewed later using evanalyzer.

pfResetEvents()

Resets the list of captured events.

Sample Use of EventView
This section provides samples of the simplest use of EventView and a more complex use.
The Simplest Case

The simplest way to use EventView is to use only OpenGL Performer internal events. In
this case, an application program may look like the following:
main()
{
pfInit();
// Initialize event capture. Call between pfInit and pfConfig.
pfInitializeEvents();
pfConfig();
// Create window & scene graph.

684

007-1680-080

EventView—A Performance Analyzer

....
// Start event capture
pfEventSampleOn();
for (i = 0 ; i < numFrames ; i ++)
{
pfSync();
pfFrame();
}
// Stop event capture.
pfEventSampleOff();
// Write to file.
pfWriteEvents("events.timefile");
}

After the program completes, you can analyze the generated events using the command
evanalyzer events.timefile.
A More Complex Case

In a more complex case, an application has some private events that it wishes to add to
the captured event file. Two stages are necessary:
Stage 1
You must define application’s private events in a file. The following example defines a
single event named myDrawCB and assigns a single information slot of type integer for
this event:
file "myEvents.timefile";
max_events 10000;
c_file "myEvents.c";
h_file "myEvents.h";
active true;
conditional_sampling true;
malloc "pfSharedMalloc";
event "myDrawCB"

007-1680-080

685

21: Performance Tuning and Debugging

{
info "CB_Info"
{
name "Something" ;
type int ;
};
};

Running the command evpp on the preceding file creates the two files myEvents.c and
myEvents.h. The application should compile and link with the newly generated file
myEvents.c. The file myEvents.h contains macros that the application uses to control
event capturing.
Stage 2
The following application code generates an event of type myDrawCB in a DRAW
callback of some node in the scene graph. Every time the program generates an event, the
program assigns a value to its information slot:
int pre_draw_cb(pfTraverser *trav, void *userData)
{
int someValue;
// Mark the start of private event MyDrawCB.
EVENT_START_MYDRAWCB;
// Assign value to the information field of this event.
EVENT_MYDRAWCB_INFO_CB_Info (someValue);
... Run callback code.
// Mark the end of private event MyDrawCB.
EVENT_END_MYDRAWCB;
}
main()
{
pfGeode *geode;
pfInit();
// Initialize Performer-internal event capture.
pfInitializeEvents();
// Initialize application-private event capture.

686

007-1680-080

EventView—A Performance Analyzer

EVENT_INIT;
pfConfig();
// Create window & scene graph.
....
// Setup DRAW callback on a node.
pfNodeTravFuncs(geode, PFTRAV_DRAW, pre_draw_cb, NULL);
// Start event capture on all initialized event sets
// (Performer-internal and application-private).
EVENT_SAMPLE_ALL_ON;
for (i = 0 ; i < numFrames ; i ++)
{
pfSync();
pfFrame();
}
// Stop event capture on all initialized event sets..
EVENT_SAMPLE_ALL_OFF;
// Write all captured events from all initialized sets to a file.
// Since we named a file in the evpp input file,
// we need not specify it here.
EVENT_WRITE_ALL;
}

As in the first example, after the program completes, you can analyze the generated
events using the command evanalyzer myEvents.timefile. This time the file
contains both OpenGL Performer internal events and application-private events
(myDrawCB). Clicking the right-hand mouse button on the line marked myDrawCB can
toggle the display of the information field on this event. Figure 21-2 shows sample
output with a user event. The top row of the display shows the user event myDrawCB.

007-1680-080

687

21: Performance Tuning and Debugging

Figure 21-2

User Event myDrawCB

Note that it was not necessary to start and stop the sampling of OpenGL Performer
internal events. The EVENT_SAMPLE_ALL_ON and EVENT_SAMPLE_ALL_OFF
macros start and stop sampling for all currently initialized event sets.

Using EventView Tools
The EventView toolset contains the following tools:
Program

Description

evpp

An event preprocessor that takes a description of user events and
generates a C program, adding the new events into an application
program.

evanalyzer

A viewing program for captured events.

evhist

A display program for generating histograms of event durations. It is
usually activated from within evanalyzer.

evgraph

A display program for plotting functions of event information slots.

evmerge

A utility program for merging event files.

The following sections describe these tools in more detail.

688

007-1680-080

EventView—A Performance Analyzer

evpp—Event Pre-Processor
The program evpp is a preprocessor for adding user events to an OpenGL Performer
program. As input, it expects a text file describing a set of new event types. It produces a
C source file and a C include file. To enable the recording of the newly generated events,
you must compile and link the new C source file in your program.
Sample Input File
The following is an example of an evpp input file:
file "out.timefile";
max_events 10000;
c_file "myEvents.c";
h_file "myEvents.h";
active true;
conditional_sampling true;
malloc "pfSharedMalloc";
group "my_APP"
0x000001;
group "my_Cull_CB" 0x000002;
group "my_APP"
{
event "my_pfFrame";
event "my_pfSync";
};
group "my_Cull_CB"
{
event "my_Cull_Group_CB"
{
size 5;
};
event "my_Cull_DCS_CB"
{
size 5;
};
};

007-1680-080

689

21: Performance Tuning and Debugging

event "myCB"
{
info "myInfo"
{
type int;
name "myInfoName";
};
}

The preceding file defines two groups of events. Each group contains two events. In the
second group (my_Cull_CB), each event is an array of five events. This is useful for
generating the same event types from different processes. In this case, each CULL
process will generate events on callbacks to pfGroups and pfDCS. The preceding file also
defines an event called myCB,which does not belong to any group. This event contains a
single information slot of type integer.
The input file contains the following directives

690

Directive

Description

file

Specifies the output filename. If the character ? is used
instead of a filename, the generated macros
EVENT_WRITE and EVENT_WRITE_ALL will require a
filename as a parameter.

max_events

Specifies the maximum number of events that will be
allocated and captured.

c_file

Specifies the filename for the source code file that evpp
generates.

h_file

Specifies the filename of the include file that evpp
generates.

active

Determines whether evpp will generate empty macros.
This directive may be true or false. This is useful when
trying to avoid generating any macros without changing
any application source code.

conditional_sampling

Determines whether events will be captured. This
directive can be true or false. If false, events will be
captured regardless of the call to EVENT_SAMPLE_ON.

007-1680-080

EventView—A Performance Analyzer

malloc

Allows the defining of an alternative memory allocation for
internal EventView use. If using anything other than
pfSharedMalloc as the malloc value, the application must
supply a function receiving an int and returning a void*
pointer to newly allocated memory. The built-in function
pfSharedMalloc() allocates memory from the
OpenGL Performer shared arena.

group

Defines a group. When defining a group of events, you must
first define a group mask. The group directive accepts a
group name and a mask. Masks may overlap. This is useful
for defining hierarchies of groups.

event

Defines single events. They may be inside or outside the
definition of a group. An event must have a name and may
have a size and information slots. In the preceding example,
an event of size 5 is effectively an array of five independent
events. This is useful for creating per-pipe events.

evpp Command-Line Options
The following are the evpp command-line options:
Option

Description

–h

Prints the usage string.

–c

Compiles the generated files. This option can make makefiles simpler.

–n

Sets all the events in the input file to inactive. The macros generated for
all events will be empty. This flag is useful when trying to disable the
instrumentation of a source file without removing the macros.

evanalyzer—EventView Graphical Analysis Program
The program evanalyzer generates a graphic display of captured events. Each line in
its display contains events of a single type. The horizontal axis of the display denotes the
time.
View of a Single Event
Figure 21-3 shows a graphic representation of a single event.

007-1680-080

691

21: Performance Tuning and Debugging

Event duration

Event end

Event start

Figure 21-3

Up-Close View of a Single Event

An event starts at a vertical bar from the bottom of the event line to its middle, continues
horizontally, and ends as a vertical bar from the middle of the event line to its top.
Keyboard and Mouse Controls
The following are the keyboard and mouse controls for evanalyzer:
Control

Description

Up/Down arrows

Changes the zoom factor of the display.

Right/Left arrows

Shifts the display window across the captured data.

Left mouse button

Picks an event start or end point. Clicking on one event start or
end point and releasing on another prints the time difference
between the two on the top of the display.

Middle mouse button Drags the event display window across the captured data.

692

Right mouse button

If the mouse is on an event line and the event has information
fields, allows toggling of their display. If the mouse is not on an
event line, this button toggles events displayed and groups
displayed.

h key

Brings up the histogram program (evhist) and runs it on the
currently selected event.

007-1680-080

EventView—A Performance Analyzer

f key

If the event line under the mouse has information slots, prompts
for a slot. Brings up a function plot program (evgraph) and runs
it on the selected information slot on the current event.

Esc key

Exits evanalyzer.

evhist—Histogram Generator
The display program evhist generates histograms of event durations. It is usually
activated from within evanalyzer.
Moving the mouse across the window prints the number of samples that are on the
histogram bar under the mouse and what time value it represents. Figure 21-4 contains
a sample screen of evhist.

Figure 21-4

007-1680-080

evhist Sample Screen

693

21: Performance Tuning and Debugging

The following are the keyboard controls for evhist:
Control

Description

Up/Down arrows

Change the resolution of the histogram.

Right/Left arrows

Change the percentage of samples on both extremes of the
histogram that we ignore (displayed as garbage on the top of the
display).

Esc key

Exits evhist.

evgraph—Event Information Plotter
The display program evgraph plots the information slot of a given event type. It is
usually activated by evanalyzer. Figure 21-5 shows a sample display of evgraph.

Figure 21-5

694

evgraph Sample Screen

007-1680-080

EventView—A Performance Analyzer

The following are the keyboard and mouse controls for evgraph:
Control

Description

Up/Down arrows

Changes the resolution of the plot.

Right/Left arrows

Translates the plot horizontally.

Middle mouse button

Drags the plot horizontally.

Understanding OpenGL Performer Internal Events
Performance tuning typically requires finding a portion of the frame time that takes too
long to execute and taking measures to shrink its duration. EventView enables bounding
some major OpenGL Performer internal functions and measuring their duration.
To make the EventView toolset useful, you have to understand the position and
relationships of the OpenGL Performer internal events and the OpenGL Performer
internal processing that they bound. This section places the OpenGL Performer internal
events in the context of OpenGL Performer functions.
pfSync() Events

An application calls the OpenGL Performer function pfSync() to synchronize with the
hardware video frame rate. OpenGL Performer packs multiple internal management
tasks into this function because the application process has already finished its
processing for this frame; so, this is the least critical time in the frame. The following
pseudo code describes the main processing units of pfSync() and shows what EventView
events are generated during the pfSync() duration:
pfSync()
{
EVENT_START - pfSync
... clean scene graph (first time).
... Recompute bounding spheres of nodes modified between the
... previous call to pfFrame and the present time.
EVENT_START - New_Updatables
...process all newly generated nodes for this frame.
EVENT_END - New_Updatables
EVENT_START - MergeBuffers
... Process call to pfMergeBuffers from a DBASE process.
... Process new nodes that the DBASE process merged.
EVENT_END - MergeBuffers

007-1680-080

695

21: Performance Tuning and Debugging

EVENT_START - App_Trav
... Run APP traversal on all nodes.
EVENT_END - App_Trav
EVENT_START - pfSync_Clean
... clean scene graph (second time)
... Recompute bounding spheres of nodes modified during the APP
... traversal.
EVENT_END - pfSync_Clean
... Update internal stats.
... Update pfASD positions.
... Update clip-texture emulation internals.
EVENT_START - pfSync_Sleep
... Sleep until start of next video frame.
EVENT_END - pfSync_Sleep
EVENT_END - pfSync
}

pfFrame Events

The OpenGL Performer function pfFrame() is responsible for initiating a new rendering
frame. It contains the minimal necessary processing for spawning other
OpenGL Performer processes for a new frame. The following pseudo code describes the
main processing units that pfFrame() contains and the EventView events it generates:
pfFrame()
{
EVENT_START - pfFrame
EVENT_START - pfFrame_Clean
... Clean scene graph.
... Recompute bounding spheres of nodes modified between the
... previous call to pfSync and the present time.
EVENT_END - pfFrame_Clean
... Check all CULL processes for completion of previous frame.
... Decide what CULL processes should start a new frame.
EVENT_START - pfFrame_Updates
... Send forked processes (CULL, ISECT, DBASE, COMPUTE) to copy
... all scene graph changes from APP copy of the scene graph to
... their private copy of the scene graph.
... Wait for all above processes to finish updating their copy
... of the scene graph.
EVENT_END - pfFrame_Updates
... Execute all unforked processes as part of the APP process.
EVENT_END - pfFrame
}

696

007-1680-080

EventView—A Performance Analyzer

CULL Process Events

The following pseudo code shows the main processing units of the CULL process and
the EventView events that it generates. The code portion shown spans a single CULL
process frame:
while(1)
{
EVENT_START - Cull
EVENT_START - Cull_Updates
... Copy all scene graph changes from the APP copy of the
... scene graph to the CULL copy of the scene graph.
EVENT_END - Cull_Updates
for each channel on this pipe do
EVENT_START - Cull_Channel
... Call CULL traversal on this channel.
EVENT_END - Cull_Channel
EVENT_END - Cull
}

DRAW Process Events

The following pseudo code shows the main processing units of the DRAW process and
the EventView events that it generates. The code portion shown spans a single DRAW
process frame:
while (1)
{
EVENT_START - Draw
... for each channel on this pipe do
EVENT_START - Draw_Channels
.... Call the drawing function on a channel.
EVENT_END - Draw_Channels
EVENT_END - Draw
}

Other Events

Each of the DBASE, COMPUTE, ISECT, and CULL_SIDEKICK processes has a single
event starting at the top of its frame and ending at its bottom. Their names are
self-explanatory.

007-1680-080

697

Chapter 22

22. Building a Visual Simulation Application Using
libpfv

This chapter describes how to use the library libpfv to build an application using a
graphical viewer.
The following sections appear:
•

“Overview” on page 699

•

“The Simplest pfvViewer Program” on page 700

•

“Adding Interaction to a pfvViewer Program” on page 701

•

“Reading XML Configuration Files” on page 702

•

“Module Scoping, Multiple Worlds and Multiple Views” on page 706

•

“Extending a pfvViewer—Writing Custom Modules” on page 709

•

“Extending a pfvViewer—Module Entry Points” on page 711

•

“Picking, Selection, and Interaction” on page 712

•

“More Sample Programs, Configuration Files, and Source Code” on page 716

Overview
OpenGL Performer includes libpfv, a C++ library for easy construction of modular,
interactive OpenGL Performer applications.
The library libpfv supports the following features:

007-1680-080

•

Reading and writing XML files

•

Specifying complex display configuration (pipes, windows, and channels) from a
file or through API calls

•

Tracking mouse and keyboard input

699

22: Building a Visual Simulation Application Using libpfv

•

Setting up user interaction with 3D scene elements

•

Managing multiple scene graphs (worlds)

•

Managing multiple camera positions (views)

•

Extending program functionality using program modules

The principal class in libpfv is pfvViewer. It allows complex multiworld and multiview
applications to be implemented in a modular fashion, allowing individual features to be
encapsulated into configurable and re-usable modules.
In addition to libpfv, OpenGL Performer includes ready-to-use modules that provide
the following features:
•

Loading geometry into a pfvViewer world

•

Picking geometry under the mouse pointer

•

Manipulating geometry (rotating, translating, scaling, deleting)

•

Navigating through a world using mouse and keyboard controls

•

Controlling the render style of models

•

Setting up colorful earth and sky backgrounds

•

Displaying 2D images in overlay

•

Saving snapshots of the rendered images

•

Smoothly transitioning from one world to another

•

Collecting and displaying scene graph statistics

The Simplest pfvViewer Program
The pfvViewer class starts with a very simple programming interface. It maintains a very
simple programming interface even when accessing high-level features. As shown in the
following program, the simplest program using pfvViewer loads a model and places the
camera at a comfortable viewing distance:
#include
#include
#include
main (int argc, char *argv[])

700

007-1680-080

Adding Interaction to a pfvViewer Program

{
// Initialize Performer
pfInit();
// Create a new pfvViewer
pfvViewer* viewer = new pfvViewer;
// Initialize loading of a model file.
pfdInitConverter(argv[1]);
// Configure/Initialize pfvViewer
viewer->config();
// Add a light source to the world.
viewer->addNode(new pfLightSource);
// Add a model to the world
viewer->addNode(pfdLoadFile(argv[1]));
// Start viewing
viewer->run();
}

Adding Interaction to a pfvViewer Program
To add interaction, load the following two standard modules into the viewer:
•

pfvmNavigator module
Allows the user to move around the 3D scene through mouse and keyboard input.

•

pfvmPicker module
Allows the user to select, manipulate, and delete portions of the 3D scene through
mouse and keyboard input.

The following program shows the addition of these modules to the simple program in
the preceding subsection:
#include
#include
#include
main (int argc, char *argv[])

007-1680-080

701

22: Building a Visual Simulation Application Using libpfv

{
pfvModule* module;
// Initialize Performer
pfInit();
// Create a new pfvViewer
pfvViewer* viewer = new pfvViewer;
// Add navigation module
module = pfvModule::load("pfvmNavigator");
viewer->addModule(module);
// Add mouse-picking module
module = pfvModule::load("pfvmPicker");
viewer->addModule(module);
// Initialize loading of model files.
pfdInitConverter(argv[1]);
// Configure/Initialize pfvViewer
viewer->config();
// Add a light source to the world.
viewer->addNode(new pfLightSource);
// Add a model to the world
viewer->addNode(pfdLoadFile(argv[1]));
// Start interaction
viewer->run();
}

Reading XML Configuration Files
A pfvViewer can read most of its parameters from an XML configuration file. A
pfvViewer configuration file, denoted by the .pfv extension, can contain any of the
following items:

702

•

Display configuration (pipes, windows, and channels)

•

Specification of multiple worlds

007-1680-080

Reading XML Configuration Files

•

Specification of multiple views (camera positions)

•

Extension modules to be loaded into the pfvViewer and their specific configuration
parameters

The simplest pfvViewer program using an XML configuration file is pfview, which is
provided as a precompiled executable with OpenGL Performer.
The source code for the pfview program looks like the following:
#include
int
main (int argc, char *argv[])
{
// Initialize Performer
pfInit();
// Create a new pfvViewer and read XML configuration file argv[1]
pfvViewer* viewer = new pfvViewer(argv[1]);
// Configure/Initialize pfvViewer.
viewer->config();
// Start interaction
viewer->run();
}

A minimal XML configuration file suitable for pfview has the following structure:

pfvmLoader

esprit.flt

pfvmTrackball

007-1680-080

703

22: Building a Visual Simulation Application Using libpfv

pfvmPicker

You can use a more complex XML configuration file, one including a display tag, to set
up complex display configurations. The following file is an example that sets up a
panoramic view over three channels, each rendered in a separate graphics pipe:

1
0

0.32,0.68,0.0,1.0
0.0,0.0,0.0
59.0,46.0

1
0

704

007-1680-080

Reading XML Configuration Files

0.64,1.0,0.0,1.0
-53.333,0.0,0.0
59.0,46.0

1
0

0.0,0.36,0.0,1.0
53.333,0.0,0.0
59.0,46.0

pfvmLoader

esprit.flt

pfvmTrackball

pfvmPicker

007-1680-080

705

22: Building a Visual Simulation Application Using libpfv

Module Scoping, Multiple Worlds and Multiple Views
In more complex pfvViewer applications, you can create multiple views and/or multiple
worlds. Each view will always render (view) one of the specified worlds. Each world
may be viewed by zero, one, or more views at any point during the life of the application.
Note: You can direct views from one world to another during the course of an
application.
The following simple XML configuration file defines two worlds, each being rendered
into a separate view:

world0

world1

view0
world0

view1
world1

706

007-1680-080

Module Scoping, Multiple Worlds and Multiple Views

myLoader
world,world0

esprit.flt

myLoader
world,world1

truck.pfb

myNavigator
view,view0

myNavigator
view,view1

007-1680-080

707

22: Building a Visual Simulation Application Using libpfv

You can achieve the same result through API calls, as shown in the following program:
#include
#include
main (int argc, char *argv[])
{
pfvModule* module;
// Initialize Performer
pfInit();
// Initialize loading of model files.
pfdInitConverter("flt");
pfdInitConverter("pfb");
// Create a new pfvViewer.
pfvViewer* viewer = new pfvViewer;
//Create first world.
pfvWorld* w0 = viewer->createWorld();
//Create second world.
pfvWorld* w1 = viewer->createWorld();
//Create first view. v0 becomes viewer’s current view.
pfvView* v0 = (pfvView*)(viewer->createView());
// Direct first view to first world.
v0->setTargetWorld(w0);
// Add navigation module to viewer’s current view (v0).
module = pfvModule::load("pfvmNavigator");
viewer->addModule(module, PFV_SCOPE_VIEW);
//Create second view. v1 becomes viewer’s current view.
pfvView* v1 = (pfvView*)(viewer->createView());
// Direct second view to second world
v1->setTargetWorld(w1);
// Add navigation module to viewer’s current view (v1).
module = pfvModule::load("pfvmNavigator");
viewer->addModule(module, PFV_SCOPE_VIEW);

708

007-1680-080

Extending a pfvViewer—Writing Custom Modules

// Configure/Initialize pfvViewer
viewer->config();
// Add car model to first world
w0->addNode(pfdLoadFile("esprit.flt"));
// Add truck model to first world
w1->addNode(pfdLoadFile("truck.pfb"));
// Start viewing
viewer->run();
}

Extending a pfvViewer—Writing Custom Modules
A pfvViewer accepts user-written modules and incorporates their functions into its
behavior. In order to extend a pfvViewer, you can write a new module. The following
very simple module informs the pfvViewer to invoke the handleEvent() method and
to print a message when the F1 key is pressed:
class myModule : public pfvModule
{
public:
myModule::myModule()
{
char keys[64];
// Declare what keyboard inputs this module is interested in.
sprintf(keys,"%c",PFVKEY_F1);
bindKeys(keys);
}
myModule::~myModule() {;}
// Keyboard event handler. pfvViewer calls this method every
// time the user hits the F1 key.
int handleEvent(int evType, char key)
{
printf("myModule::handleEvent called for key %s\n",
pfvInputMngr::getKeyName(key) );

007-1680-080

709

22: Building a Visual Simulation Application Using libpfv

return 0;
}
};

In order to add this module to a pfvViewer program, add the following line:
viewer->addModule(new myModule);

Note that if this module is scoped to a view, pfvViewer will only inform the module of
F1 key presses within channels belonging to that specific view. Similarly, if this module
is scoped to a specific world, pfvViewer would inform the module of F1 key presses over
any channel belonging to any view currently viewing such world.
You can scope this module to a world by making the following call:
pfvWorld* w;
viewer->addModule(new myModule, PFV_SCOPE_WORLD);

You can scope this module to a view by making the following call:
pfvView* w;
viewer->addModule(new myModule, PFV_SCOPE_VIEW);

The following example illustrates how to implement a basic custom module that controls
the camera position (for the first view in pfvViewer’s list) based on the mouse position:
class myModule : public pfvModule
{
public:
myModule(){
bindCallback(PFV_CB_FRAME);
}
~myModule(){;}
void frame() {
// Only set eye for view0 if mouse is over view0
if(pfvInputMngr::getFocusViewIndex()!=0)
return;
pfVec3 xyz,hpr;
// Get current eye position (we don't want to change xyz)
viewer->getView(0)->getEye(xyz,hpr);
float mx, my;

710

007-1680-080

Extending a pfvViewer—Module Entry Points

// Get current mouse position in view-normalized values
// (0.0 to 1.0)
pfvInputMngr::getViewNormXY( &mx, &my );
// Compute new values for Heading and Pitch based on mouse
// position
hpr[0]= (mx-0.5f)*180.0f;
hpr[1]= (my-0.5f)*-90.0f;
// Set new eye position for view0
viewer->getView(0)->setEye(xyz,hpr);
}
};

Extending a pfvViewer—Module Entry Points
A module can gain program control at the various stages of rendering. The following are
some of these stages:
•

Event-Driven methods
handleEvent()
Called in response to a key-press event. A pfvViewer invokes this method only if
the pressed key was bound by this module and if the event was generated over a
view relevant to the module.

•

Configuration methods (called once in the life of the application)
–

preConfig()
Called before pfvViewer calls pfConfig().

–

postConfig()
Called after the pfvViewer calls pfConfig().

•

Run-Time methods (called every frame):
–

sync()
Called each frame immediately after pfvViewer calls pfSync().

–

frame()
Called each frame after pfvViewer calls pfFrame().

007-1680-080

711

22: Building a Visual Simulation Application Using libpfv

–

preCull()
Called in all CULL processes before calling pfCull().

–

postCull()
Called in all CULL processes after calling pfCull().

–

preDraw()
Called in all DRAW processes before calling pfDraw().

–

postDraw()
Called in all DRAW processes after calling pfDraw().

–

overlay()
Called in all DRAW processes after postDraw() callbacks.

•

Enter and exit methods (called for scoped modules only)
–

enterWorld()
A pfvViewer invokes this method on view-scoped modules to inform them that
their view is about to start viewing a new world.

–

exitWorld()
A pfvViewer invokes this method on view-scoped modules to inform them that
their view is about to stop viewing current world.

–

enterView()
A pfvViewer invokes this method on world-scoped modules to inform them
that a new view is about to start viewing their world.

–

exitView()
A pfvViewer invokes this method on world-scoped modules to inform them
that a view is about to stop viewing their world.

Picking, Selection, and Interaction
The library libpfv also provides a framework for specifying custom interaction
behavior through the pfvPicker, pfvInteractor, and pfvSelector classes.

712

007-1680-080

Picking, Selection, and Interaction

A single pfvPicker instance will be able to coordinate multiple interaction classes derived
from pfvInteractor and/or pfvSelector.
The following example illustrates how to derive entities from the pfvInteractor class in
order to be able to highlight the geometry under the mouse cursor:
#include
#include
#include
#include
#include
#include

class myInteractor : public pfvInteractor
{
public:
myInteractor(){
// create and configure a pfHighlight instance
hl = new pfHighlight;
hl->setMode( PFHL_LINES );
hl->setColor( PFHL_FGCOLOR, 1.0f, 1.0f, 0.0f );
}
~myInteractor(){ pfDelete(hl); }
// startHlite will be called by picker once whenever mouse cursor
// is moved over some geometry after being pointed away from all
// geometry.
int startHlite( pfvPicker*p, int prmsn ){
// Obtain a pointer to the pfNode that was picked by picker
p->getPickResults(&node);
// Traverse picked node and highlight it
pfuTravNodeHlight( node, hl );
// return 1 indicating we accept highlighted state
return 1;
}
// updateHlite will be called by picker on each frame as long as
// mouse cursor remains over some geometry.
int updateHlite( pfvPicker* p,int ev,int prmsn, pfvPickerRequest*r
){
pfNode* curnode = node;
// Obtain a pointer to the pfNode that was picked by picker
p->getPickResults(&node);

007-1680-080

713

22: Building a Visual Simulation Application Using libpfv

// if node picked by picker is not the node that is
// currently highlighted
if(node!=curnode)
{
// De-highlight previously highlighted node
pfuTravNodeHlight( node, hl );
// Traverse picked node and highlight it
pfuTravNodeHlight( curnode, NULL );
}
return 1;
}
// endHlite will be called by picker once whenever mouse cursor is
// moved away from all geometry..
void endHlite( pfvPicker* p ){
// De-highlight previously highlighted node
pfuTravNodeHlight( node, NULL );
}
private:
pfHighlight* hl;
pfNode *node;
};

class myModule : public pfvModule
{
public:
myModule(){
// This module will use two callbacks (entry-points):
bindCallback(PFV_CB_POSTCONFIG);
bindCallback(PFV_CB_FRAME);
}
~myModule(){;}
// postConfig is called once by pfvViewer, after calling
// pfConfig().
void postConfig(){
// Create a pfvInputMngrPicker instance
picker = new pfvInputMngrPicker;
// Set up picker so it will automatically isect scene and
// allow interactors to highlight

714

007-1680-080

Picking, Selection, and Interaction

picker->setState(PFPICKER_ALLOW_HLITE, NULL, NULL);
// Create an instance of our custom interactor
ia = new myInteractor;
// set up a pointer to our interactor on scene's root-node
ia->nodeSetup( viewer->getWorld(0)->getScene(), picker);
}
// on every frame, call picker->update() from APP process.
void frame() {
picker->update();
}
private:
pfvInputMngrPicker* picker;
myInteractor* ia;
};
int
main (int argc, char *argv[])
{
pfInit();
pfvViewer* viewer = new pfvViewer();
pfFilePath(".:/usr/share/Performer/data");
pfdInitConverter("esprit.flt");
viewer->addModule(new myModule);
viewer->addModule(pfvModule::load("pfvmTrackball"));
viewer->config();
viewer->addNode(new pfLightSource);
viewer->addNode(pfdLoadFile("esprit.flt"));
viewer->run();
}

007-1680-080

715

22: Building a Visual Simulation Application Using libpfv

More Sample Programs, Configuration Files, and Source Code
OpenGL Performer provides many libpfv sample programs, configuration files, and
source code for modules in the following directories:
•

/usr/share/Performer/src/pguide/libpfv/picker (IRIX and Linux)
%PFROOT%/Src/pguide/libpfv/picker (Microsoft Windows)
The samples in this directory demonstrate the use of pfvPicker and derived classes
as well as how to extend the pfvInteractor and pfvSelector classes to implement
your custom interaction behaviors.

•

/usr/share/Performer/src/pguide/libpfv/viewer (IRIX and Linux)
%PFROOT%/Src/pguide/libpfv/viewer (Microsoft Windows)
The samples in this directory demonstrate how to do the following:

•

–

Load models into pfvViewer applications.

–

Load standard or custom modules.

–

Create pfvViewers with multiple camera positions (views).

–

Create pfvViewers with multiple independent scene graphs (worlds).

–

Create complex multichannel display configurations.

–

Load complex display configurations from an XML file.

–

Write custom modules.

–

Compile modules into re-usable DSOs.

/usr/share/Performer/src/pguide/libpfv/viewer/modules
(IRIX and Linux)
%PFROOT%/Src/pguide/libpfv/viewer/modules (Microsoft Windows)
This directory contains source code for the following modules:

716

–

pfvmDrawStyle

–

pfvmEarthSky

–

pfvmLoader

–

pfvmLogo

–

pfvmNavigator

–

pfvmPicker

007-1680-080

More Sample Programs, Configuration Files, and Source Code

•

–

pfvmSnapshot

–

pfvmStats

–

pfvmTrackball

–

pfvmWorldSwitcher

/usr/share/Performer/config (IRIX and Linux)
%PFROOT%/Config (Microsoft Windows)
This directory contains examples of pfvViewer configuration files, denoted by the
.pfv filename extension.

007-1680-080

717

Chapter 23

23. Programming with C++

This chapter provides an overview of some of the differences between programming
OpenGL Performer using the C++ Application Programming Interface (C++ API) rather
than the C-language Application Programming Interface (C API), which is described in
the earlier chapters of this guide.

Overview
Although this guide uses the C API throughout, the C++ API is in every way equal and
in some cases superior in functionality and performance to the C API.
Every function available in the C API is available in the C++ API. All of the C API
routines tightly associated with a class have a corresponding member function in the
C++ API; for example, pfGetDCSMat() becomes pfDCS::getMat(). Routines not closely
associated with a class are the same in both APIs. Examples include high-level global
functions such as pfMultiprocess() and pfFrame() and low-level graphics functions such
as pfAntialias().
Most of the routines associated with a class can be divided into three categories: setting
an attribute, getting attribute, and acting on the object. In the C API, sets were usually
expressed as pf, gets as pfGet and simple actions
as pf, where is the abbreviation for the full name of the class.
In some cases where there was no room for confusion or this usage was awkward, the
routine names were shortened, for example, pfAddChild().
The principal difference in the naming of member functions in the C++ API and the
corresponding routine name in the C-language API is in the naming of member functions
where the “pf” prefix and the identifier are dropped. In addition, the word “set”
or “get” is prefixed when attribute values are being set or retrieved. Hence, value setting
functions are usually have names of the form pfClass::set, value getting
functions are named pfClass::get, and actions appear as pfClass::.

007-1680-080

719

23: Programming with C++

Table 23-1

Corresponding Routines in the C and C++ API

C Routine

C++ Member Function

Description

pfMtlColor()

pfMaterial::setColor()

Set material color.

pfGetMtlColor()

pfMaterial::getColor()

Get material color.

pfApplyMtl()

pfMaterial::apply()

Apply the material.

Note: Member function whose names begin with “pf_”, “pr_” or “nb_” are internal
functions and should not be used by applications. These functions may have
unpredictable side effects and also should not be overridden by application subclasses.

Class Taxonomy
There are three main types of C++ classes in OpenGL Performer. The following
description is based on this categorization of the main types: public structs, libpr
classes, and libpf classes. A fourth distinct class is pfType, the class used to represent
the type of libpr and libpf classes.

Public Structs
These classes are public structs with exposed data members. They include pfVec2,
pfVec3, pfVec4, pfMatrix, pfQuat, pfSeg, pfSphere, pfBox, pfCylinder, and pfSegSet.

libpr Classes
These classes derive from pfMemory. When multiprocessing, all processes share the
same copy of the object’s data members.

720

007-1680-080

Programming Basics

libpf Classes
These classes derive from pfUpdatable and when multiprocessing, each APP, CULL, and
ISECT process has a unique copy of the object’s data members.

pfType Class
As with the C API, information about the class hierarchy is maintained with pfType
objects.

Programming Basics
Header Files
For IRIX and Linux, the C++ include files for libpf and libpr are in
/usr/include/Performer/pf and /usr/include/Performer/pr, respectively.
For Microsoft Windows, the corresponding header files are in %PFROOT%/Include/pf
and %PFROOT%/Include/pr. An application using a class should include the
corresponding header file.
Table 23-2

libpf

007-1680-080

Class

Header Files for libpf Scene Graph Node Classes
Include File

pfASD

pfBillboard

pfDoubleDCS

pfDoubleFCS

pfDoubleSCS

pfDCS

pfFCS

pfGeode

721

23: Programming with C++

Table 23-2

libpf

Class

Include File

pfGroup

pfIBRnode

pfLOD

pfLayer

pfLightPoint

pfLightSource

pfNode

pfPartition

pfSCS

pfScene

pfSequence

pfSwitch

pfText

Table 23-3

722

Header Files for libpf Scene Graph Node Classes (continued)

Header Files for Other libpf Classes

libpf Class

Include File

pfBuffer

pfChannel

pfCullProgram

pfEarthSky

pfLODState

pfMPClipTexture

pfPipe

007-1680-080

Programming Basics

Table 23-3

Header Files for Other libpf Classes (continued)

libpf Class

Include File

pfPipeWindow

pfRotorWash

pfShader

pfShadow

pfShaderManager

pfPipeVideoChannel
pfTraverser
pfPath

pfVolFog

Table 23-4

libpr

007-1680-080

Class

Header Files for libpr Graphics Classes
Include File

pfColortable

pfClipTexture

pfDispList

pfFog

pfFont

pfFBState

pfGeoSet
pfHit

pfGeoState

pfHighlight

pfIBRtexture

pfPassList

723

23: Programming with C++

Table 23-4

libpr

Class

Include File

pfLPointState

pfLight
pfLightModel

pfMaterial

pfSprite

pfState

pfString

pfTexture
pfTexGen
pfTexEnv

Table 23-5

724

Header Files for libpr Graphics Classes (continued)

Header Files for Other libpr Classes

libpr Class

Include File

pfCycleBuffer
pfCycleMemory

pfDataPool

pfEngine

pfFile

pfFlux

pfSphere
pfBox
pfCylinder
pfPolytope
pfFrustum
pfSeg
pfSegSet

007-1680-080

Programming Basics

Table 23-5

Header Files for Other libpr Classes (continued)

libpr Class

Include File

pfVec2
pfVec3
pfVec4
pfMatrix
pfQuat
pfMatStack

pfList

pfMemory

pfObject

pfQueue

pfStats

There are additional C++ object classes in the utility libraries. Header files for those
classes are similarly named with their own library name for the directory and prefix for
the header file name. The libpui C++ has a full C++ API and its header files are named
like the example . The libpfutil library has
some C++ classes and the header files are named as
.

Creating and Deleting OpenGL Performer Objects
The OpenGL Performer base classes all provide operator new and operator delete. All
libpr and libpf objects, except pfObject, pfMemory, and their derivatives, must be
explicitly created with operator new and deleted with operator delete. Objects of these
classes cannot be created statically on the stack or in arrays.
All objects of classes derived from pfObject or pfMemory are reference counted and must
be deleted using pfDelete(), rather than the delete operator. pfDelete() checks the

007-1680-080

725

23: Programming with C++

reference count of the object and, when multiprocessing, delays the actual deletion until
other processes are done with the object. To decrement the reference count and delete
with a single call use pfUnrefDelete().
Note: Public structs such as pfVec3, pfSphere, etc. may be deleted either with pfDelete()
or the delete operator.
The default new operator creates objects in the current shared memory arena if one
exists. libpr objects and public structures have an additional new operator that takes
an arena argument. This new operator allows allocation from the heap (indicated by an
arena of NULL) or from a shared memory arena created by the application with
acreate().
Example 23-1

Valid Creation of Objects in C++

// valid creation of libpf objects
pfDCS
*dcs = new pfDCS;
// only way
// valid creation of libpr objects
pfGeoSet *gs = new pfGeoSet;
// from default arena
pfGeoSet *gs = new(NULL) pfGeoSet; // from heap
// valid creation of public structs
pfVec3 *vert = new pfVec3;
// from default arena
pfVec3 *verts = new pfVec3[10];
// array from default
static pfVec3 vert(0.0f, 0.0f, 0.0f); // static
Example 23-2

Invalid Creation of Objects in C++

// invalid creation of libpf objects
pfDCS
*dcs = new(NULL) pfDCS;
// not in shared mem
pfDCS
*dcs = new pfDCS[10];
// array
// invalid creation of libpr objects
pfGeoSet *gs = new pfGeoSet[10];
// array
// invalid creation of public structs
pfVec3 *vert = new(NULL) pfVec3[10];// array, non-default new

726

007-1680-080

Porting from C API to C++ API

Caution: This last item in Example 23-2 is invalid because C++ does not provide a
mechanism to delete arrays of objects allocated with a new operator defined to take
additional arguments, for example, operator new(size_ts, void *arena). Attempting to
delete an array of objects allocated in this manner can cause unpredictable and fatal
results such as the invocation of the destructor a large number of times on pointers inside
and outside of the original allocation.

Invoking Methods on OpenGL Performer Objects
Since libpr and libpf objects are allocated, they can only be maintained by reference.

Passing Vectors and Matrices to Other Libraries
Passing arrays of floats is very common in graphics programming. Calls to OpenGL
often require an array of floats or a matrix. In the C API, the data types such as pfMatrix
are arrays and so can be passed straight through to OpenGL routines; the following is an
example:
pfMatrix ident;
pfMakeIdentMat(ident);
glLoadMatrix(ident);

In the C++ API, the data field of the pfMatrix must be passed instead, as in this example:
pfMatrix ident;
ident.makeIdent();
glLoadMatrix(ident.mat);

Porting from C API to C++ API
When compiled with C++, OpenGL Performer supports three usages of the API:
1.

Pure C++ API. This is the default style of usage.

2. Pure C API. This can be achieved by defining the token PF_CPLUSPLUS_API to be
0, for example, by adding the following line in source files before they include any
OpenGL Performer header files:
#define PF_CPLUSPLUS_API 0

007-1680-080

727

23: Programming with C++

In this mode all data types are the same as when compiling with C.
3. C++ API and C API. This mode can be enabled by defining the token PF_C_API to
be 1, for example, by adding the following line in source files before they include
any OpenGL Performer header files:
#define PF_C_API 1

In this mode, both C++ and C functions are available and data types are C++. See
the section below concerning passing certain data types.

Typedefed Arrays Versus Structs
In the C API, the pfVec2, pfVec3, pfVec4, pfMatrix and pfQuat data types are all
typedefed arrays. In the C++ API, they are all structs. When converting C code to use the
C++ API or when compiling C API code with both APIs enabled, be sure to change
routines in your code that pass objects of these types. In the C++ API, you almost always
want to pass arguments of these types by reference rather than by value.
For example, the following C API routine should be rewritten for the C++ API to pass by
reference:
void MyVectorAdd(pfVec2 dst, pfVec2 v1, pfVec2 v2)
{
dst[0] = v1[0] + v2[0];
dst[1] = v1[1] + v2[1];
}

Two possible alternatives follow:
void MyVectorAdd(pfVec2& dst, pfVec2& v1, pfVec2& v2)
{
dst[0] = v1[0] + v2[0];
dst[1] = v1[1] + v2[1];
}

or
void MyVectorAdd(pfVec2* dst, pfVec2* v1, pfVec2* v2)
{
dst->vec[0] = v1->vec[0] + v2->vec[0];
dst->vec[1] = v1->vec[1] + v2->vec[1];
}

728

007-1680-080

Subclassing pfObjects

Without this change, time will be wasted copying v1 and v2 by value and the result will
not be returned to the routine calling MyVectorAdd().

Interface Between C and C++ API Code
The same difference in passing conventions applies if you are calling a C function from
code that uses the C++ API. Functions passing typedefed arrays with the C API must
have a different prototype for use with the C++ API. Macros for use in C prototypes
bilingual can be found in /usr/include/Performer/prmath.h for IRIX and Linux
and in %PFROOT%/Include/prmath.h for Microsoft Windows.
#if PF_CPLUSPLUS_API
#define PFVEC2 pfVec2&
#define PFVEC3 pfVec3&
#define PFVEC4 pfVec4&
#define PFQUAT pfQuat&
#define PFMATRIX pfMatrix&
#else
#define PFVEC2 pfVec2
#define PFVEC3 pfVec3
#define PFVEC4 pfVec4
#define PFQUAT pfQuat
#define PFMATRIX pfMatrix
#endif /* PF_CPLUSPLUS_API */

These macros are used in the C API prototypes for OpenGL Performer that pass
typedefed arrays, as shown in the following example:
extern float pfDotVec2(const PFVEC2 v1, const PFVEC2 v2);

But they are not necessary or appropriate for when passing pointers to typedefed arrays
in C, because a pointer to a struct is passed in the same manner as a pointer to an array.
This is shown in the following example:
extern void pfFontCharSpacing(pfFont *font, int ascii,
pfVec3 *spacing);

Subclassing pfObjects
With the C API, the main mechanism for extending the functionality of the classes
provided in OpenGL Performer is the specification of the user data pointer on pfObjects

007-1680-080

729

23: Programming with C++

with pfUserData() and the specification of callbacks on pfNodes with
pfNodeTravFuncs() and pfNodeTravData(). The C++ API also supports these
mechanisms, but also provides the additional capacity to subclass new data types from
the classes defined in OpenGL Performer. Subclassing allows additional member data
fields and functions to be added to OpenGL Performer classes. At its simplest,
subclassing merely provides a way of adding additional data fields that is more elegant
than hanging new data structures off of a pfObject’s user data pointer. But in some uses,
subclassing also allows significantly more control over the functional behavior of the
new object because virtual functions can be overloaded to bypass, replace, or augment
the processing handled by the parent class from OpenGL Performer.

Initialization and Type Definition
The new object should provide two static functions, a constructor that initializes the
instances pfType*, and a static data member for the type system as shown in the
following table:
Table 23-6

Data and Functions Provided by User Subclasses

Class Data or Function

Function

static void init()

Initializes the new class.

static pfType* getClassType()

Returns the pfType* of the new class.

static pfType* classType

Stores the pfType* of the new class.

constructor

Sets the pfType* for each instance.

The init() member function should initialize any data structures that are related to the
class as a whole, as opposed to any particular instance. The most important of these is
the entry of the class into the type system. For example, the Rotor class defined in the
Open Inventor loader (see Rotor.h and Rotor.C in
/usr/share/Performer/src/lib/libpfdb/libpfiv for IRIX and Linux and in
%PFROOT%/Src/lib/libpfdb/libpfiv for Microsoft Windows) is a subclass of
pfDCS. Its initialization function merely enters the class into the type system.
Example 23-3

Class Definition for a Subclass of pfDCS

public Rotor : public pfDCS
{
static void init();

730

007-1680-080

Subclassing pfObjects

static pfType* getClassType(){ return classType; };
static pfType* classType;
}
pfType *Rotor::classType = NULL;
Rotor::Rotor()
{
setType(classType);
...
}

// set the type of this instance

void
Rotor::init()
{
if (classType == NULL)
{
pfDCS::init();
classType =
new pfType(pfDCS::getClassType(), “Rotor”);
}
}

As described in the following section, the initialization function, Rotor::init() should be
called before pfConfig().

Defining Virtual Functions
Below is the example of the Rotor class, which specifies the traversal function for the
libpf application traversal. When overloading a traversal function, it is usually
desirable to invoke the parent class function, in this case, pfDCS::app(). It is not currently
possible to overload libpf’s intersection or culling traversals. See “Multiprocessing and
libpf Objects” on page 735.
Example 23-4

Overloading the libpf Application Traversal

int
Rotor::app(pfTraverser *trav)
{
if (enable)
{
pfMatrix mat;
double now = pfGetFrameTimeStamp();

007-1680-080

731

23: Programming with C++

// use delta and renorm for large times
prevAngle += (now - prevTime)*360.0f*frequency;
if (prevAngle > 360.0f)
prevAngle -= 360.0f;
mat.makeRot(prevAngle, axis[0], axis[1], axis[2]);
setMat(mat);
prevTime = now;
}
return pfDCS::app(trav);
}
int
Rotor::needsApp(void)
{
return TRUE;
}

The same behavior could also be implemented in either the C or C++ OpenGL Performer
API using a callback function specified with pfNodeTravFuncs().
Note: Classes of pfNodes that need to be visited during the application traversal even in
the absence of any application callbacks should define the virtual function needsApp()
to return TRUE.

Accessing Parent Class Data Members
Accesses to parent class data is made through the functions on the parent class. Data
members on built-in classes should never be accessed directly.

Multiprocessing and Shared Memory
Note: This section does not pertain to Microsoft Windows.

732

007-1680-080

Multiprocessing and Shared Memory

Initializing Shared Memory
In general, to assure safe multiprocess operation with any DSOs providing C++ virtual
functions or defining new pfTypes, initialization should be carried out in the following
sequence:
1.

Call pfInit(). This initializes the type system and, for libpf applications, sets up
shared memory.

2. Call the init function for utility libraries that you are using if you use their C++ API.
This includes pfuInit() for libpfutil and pfiInit() for libpfui.
3. Initialize any application-supplied classes:
a.

Load any application-specific C++ DSOs.

b. Call pfdInitConverter() to initialize and load any converter DSOs and allow
those DSOs to initialize any potential C++ classes.
c.

Enter any user-supplied pfTypes into the type system; for example, call this
function:
Rotor::init()

4. Call pfConfig(). This forks off other processes as specified by pfMultiprocess().
New pfTypes created after this point cannot be used in any forked processes.
5. Create libpf and libpr objects.
Note: Pure libpr applications that do their own multiprocessing outside of OpenGL
Performer with fork() should explicitly create shared memory with pfInitArenas()
before calling pfInit(). Otherwise, the type system will not be visible in the address space
of other processes.

More on Shared Memory and the Type System

Note: This is an advanced section.
OpenGL Performer objects or other objects that use pfTypes can only be shared between
related processes. Related processes are those created with fork() or sproc() from the
main process after pfInit() in a libpf application, for example, processes created by
pfConfig().

007-1680-080

733

23: Programming with C++

New pfTypes should be added before pfConfig() forks off other processes so that the
static data member containing the class type is visible in all processes, otherwise
pf::getClassType() will return NULL in other processes. This effectively
precludes the creation of subclasses of OpenGL Performer objects after pfConfig().
Virtual Address Spaces and Virtual Functions

Note: This is an advanced section.
When using virtual functions, it is very important that the object code reside at the same
address in all processes. Normally, this is not an issue since the object code for all
OpenGL Performer classes is loaded (whether statically linked or loaded as dynamic
shared objects, DSOs) before pfConfig() is called to fork off processes. For user-defined
C++ classes with virtual functions, it is important that the object code reside at the same
virtual address space in all processes that access them. For this reason, the DSOs for any
user-defined classes should be loaded before pfConfig(), regardless of whether they use
the pfType system or not.

Data Members and Shared Memory
Non-static Member Data

The default operator new for objects derived from pfObject causes all instances to be
created in shared memory so that objects will be visible to other related processes that
need to see them.
Static Member Data

Classes having static data members that may change value and need to be visible from
all processes should allocate shared memory for the data (for example, pfMalloc or new
pfMemory) and set the static data member to point to this memory before pfConfig() as
shown in the following example.
Example 23-5

Changeable Static Data Member

class Rotor : public pfDCS
{
static int* instanceCount;

734

007-1680-080

Multiprocessing and Shared Memory

}
Rotor::instanceCount = NULL;
void Rotor::init()
{
...
instanceCount = new(sizeof(int)) pfMemory;
*instanceCount = 0;
}
Rotor::Rotor()
{
...
(*instanceCount)++; // increment the creation counter
}

A static data member whose value is set before pfConfig() and never changes thereafter
does not need to be allocated from shared memory. The classType member of Rotor is an
example of this since the class should be initialized; that is, call Rotor::init() before
pfConfig().

Multiprocessing and libpf Objects
Note: This is an advanced topic.
The multiprocessing behavior of libpf objects (that is, those deriving from pfNode or
pfUpdatable) differs from that of libpr objects. Both are typically created in shared
memory but, with a libpr object, all processes share the same data members while
libpf objects have a built-in multiprocessing data mechanism that provides different
copies in the APP, CULL, and ISECT stages of the OpenGL Performer pipeline. The term
multibuffering refers to the maintenance and frame-accurate updating of these data.
With a user-defined subclass of a libpf class, the original data elements of the libpf
parent class are still multibuffered. However, the parallel multibuffer copies maintained
in the other processes are instances of the parent class rather than the subclass. This is not
normally visible to the application, since even for callbacks in the CULL and ISECT
processes, the application always works from the pointer to the copy used in the APP
process, in part, so that objects can be identified by comparison of pointers. However, this
difference would be visible if the virtual traversal functions for culling or intersection
were overloaded. These virtual functions should not be overloaded by the subclass since

007-1680-080

735

23: Programming with C++

they will not have any effect when the CULL or ISECT stages are in separate processes.
Node callbacks specified with pfNodeTravFuncs() should be used instead.
If you require multibuffering of your subclassed data members, use a pfCycleBuffer or a
pfFlux to hold this data.

Performance Hints
Constructor Overhead
It is quite natural to frequently construct and destroy arrays of public structs such as
pfVec3 on the stack. Beware, even though the constructors for these classes are empty, it
still requires a function call for each element of the array. The same applies to classes that
contain arrays of structs; for example, pfSegSet contains an array of pfSegs.

Math Operators
Assignment operators, for example, “+=”, are significantly faster than their
corresponding binary operators, for example, “+”, because the latter involves
constructing a temporary object for the return value.

736

007-1680-080

Glossary

alias
An alternate extension for a file type as processed by the pfdLoadFile() utility. For
example, VRML “.wrl” files are in a sense an alias for Open Inventor “.iv” files since
the Open Inventor loader can read VRML files as well. Once the alias is established, files
with alternate extensions will be loaded by the designated loader.
application buffer
The main (and usually only) buffer of libpf data structures such as the nodes in the
scene graph. Alternate buffers may be created and data can be constructed in these new
buffers from parallel processes to support high-performance asynchronous database
paging during real-time simulation.
arena
An area (allocation area) from which shared memory is allocated. Usually the arena is the
default one created by pfInit() or pfInitArenas(), but some objects (for example, those in
libpr) may be created in any arena returned by acreate(). OpenGL Performer calls that
accept an arena pointer as an argument can also accept the NULL pointer, indicating that
the memory should be allocated from the heap. See also heap.
asynchronous database paging
Asynchronous database paging, an advanced method of scene-graph creation, allows
desired data to be read from a disk or network connection and OpenGL Performer
internal data structures to be built for this data using one or more processes running on
separate CPUs rather than performing these tasks in the application process. Once the
data structures are created in these database processes, they must be explicitly merged
into the application buffer.
attribute binding
The binding of an attribute specifies how often an attribute is specified and the scope of
each specification. For example, given a collection of triangles for rendering, a color can
be specified with each vertex of each triangle, with each triangle, or once for the entire
collection of triangles.

007-1680-080

737

Glossary

base geometry
The object with the lowest visual priority in a pfLayer node’s list of children. This would
be the runway, for example, in a runway and stripes airport database example. See also
layer geometry.
bezel
The beveled border region surrounding any item, but most notably, around the edge of
a CRT monitor.
billboard
Geometry that rotates to follow the eyepoint. This is often simply a single
texture-mapped quadrilateral used to represent an object that has roughly cylindrical or
spherical symmetry, such as a tree or a puff of smoke, respectively. OpenGL Performer
supports billboards that can rotate about an axis for cylindrical objects or a point for
spherical objects.
binning
The action of the sort phase of libpf’s cull traversal that segregates drawable geometry
into major sections (such as opaque and transparent) before the per-bin sorting based on
the contents of associated pfGeoStates so that state changes can be reduced by drawing
groups of similar geometry sequentially while still drawing semitransparent objects in
the desired order within the frame.
bins
The unique collections into which the cull traversal segregates drawable geometry. The
number of bins is defined by calls to pfChanBinSort() and pfChanBinOrder(). Typical
bins are those for opaque and transparent geometry, where opaque objects are rendered
first for superior image quality when using blended transparency.
blur margin
A parameterized value used to control blurring and flickering of textures.
bounding volume
A convex region that encompasses a geometric object or a collection of such objects.
OpenGL Performer pfGeoSets have axis-aligned bounding boxes, which are rectangular
boxes whose faces are along the X, Y, or Z axes. Each OpenGL Performer pfGeode has a
bounding sphere that contains the bounding box of each pfGeoSet in the pfGeode.
OpenGL Performer group nodes have hierarchical bounding spheres that contain
(bound) the geometry in their descendent nodes. The purpose of bounding volumes is to

738

007-1680-080

Glossary

allow a quick test of a region for being off-screen or out of range of intersection search
vectors.
buffer scope
All OpenGL Performer nodes are created in a pfBuffer, either the primary application
buffer or in an alternate returned by pfNewBuffer(). Only one pfBuffer is considered
“current,” and this buffer can be selected using pfSelectBuffer(). All new nodes are
created in the current pfBuffer and will be visible only in the current buffer until that
pfBuffer is merged into the application buffer using pfMergeBuffer(). A process cannot
access nodes that do not have scope in its current buffer except through the special
“buffer” commands: pfBufferAddChild(), pfBufferRemoveChild(), and
pfBufferClone(). Thus, they are said to have buffer scope.
channel
A visual channel specifies how a geometric scene should be rendered to the display
device. This includes the viewport area on the screen as well as the location, orientation,
and field of view associated with the viewer or camera.
channel group
A set of channels that share attributes such as the eyepoint or callbacks. When a shared
attribute is set on any member of the group, all members get the new value. Channel
groups are most commonly used for adjacent displays making up a panorama.
channel share mask
A bit mask indicating the attributes that are shared by all channels in a channel group.
Typical shared attributes are field of view, view specification, near and far clipping
distances, the scene to be drawn, stress parameters, level of detail parameters, the
earth/sky model, and swapbuffer timing.
children
OpenGL Performer’s hierarchical scene graph of pfNodes has internal nodes derived
from the pfGroup class, and each node attached below a pfGroup type node is known as
a child of that node. The complete list of child nodes are collectively termed the children
of that node.
class hierarchy
The source through which OpenGL Performer classes are defined. This class hierarchy
defines the data elements and member functions of these data types through the notion
of class inheritance.

007-1680-080

739

Glossary

class inheritance
Class inheritance describes the process of defining one object as a special version of
another. For example, a pfSwitch node is a special version of a pfGroup node in that it
has control information about which children are active for drawing or intersection. In
all other respects, a pfSwitch has the same capabilities as a pfGroup, and the OpenGL
Performer API supports this notion directly in both the C and C++ API by allowing a
pfSwitch node to be used wherever a pfGroup is called for in a function argument. This
same flexibility is supported for all derived types.
clipped
Geometry is said to be clipped when some or all of its geometric extent crosses one or
more clipping planes and the portion of the geometry beyond the clipping plane is
mathematically trimmed and discarded.
clipping planes
The normal clipping planes are those that define the viewing frustum. These are the left,
right, top, bottom, near, and far clipping planes. All rendered geometry is clipped to the
intersection of the half-spaces defined by these planes and only the portion inside all six
is displayed by the graphics hardware.
clip texture
This entity virtualizes MIPmapped textures using hardware and software support so
that only the texels in the region close to the viewer (known as the clipped region) need
to be loaded in texture memory. Also known as ClipMaP.
cloned instancing
The style of instancing that creates a (possibly partial) copy of a node hierarchy rather
than simply making a reference to the parent node. This allows pfDCS nodes and other
internal nodes to be changed in the copy without changing those in the original. Also see
shared instancing.
cloning
Making a copy of a data structure recursively copying down to some specified level. In
OpenGL Performer pfCopy() creates a shallow copy. pfClone() creates a deeper copy
that creates new copies of internal nodes but not of leaf nodes. This means that the
pfDCS, pfSwitch, and other internal nodes in the cloned hierarchy are separate from
those in the original.

740

007-1680-080

Glossary

compiled mode
OpenGL Performer pfGeoSets are designed for rapid immediate-mode rendering and in
most situations outperform OpenGL display list usage. In those cases where GL display
lists are desired, pfGeoSets may be placed in compiled mode, whereby a GL display list
will be created the first time the pfGeoSet is rendered and this display list will be used
for subsequent renderings until the pfGeoSet compiled mode flag is explicitly reset. Once
a pfGeoSet is compiled, any changes to its data arrays by the pfMorph node or other
means will not be effective until the compiled-mode flag is cleared.
complex pixels
Pixels for which several different geometric primitives contribute to the pixel’s assigned
color value. Such pixels are rare in typical scenes and only exist at edges of polygons
unless multisample blending is in use. When this blending mode is used, all pixels
rendered as neither fully opaque nor fully transparent are complex pixels.
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.
critically damped
A closed-loop control system notion where the feedback transfer function is just right:
not so slow that the system goes out of range before correction is applied and not so fast
that overcorrection causes rapid swings or variation. This should be the goal of any
user-specified stress management function.
cull
See culling.
cull volume visualization
The visual display of the culling volume, usually the same as the viewing frustum, to
which the scene is culled before rendering. Normally the projected culling volume fills
the display area. By rendering with a larger field of view or from an eyepoint that differs
from the origin of the frustum, the tightness of culling can be determined for database
tuning. The culling volume itself is often drawn in wireframe.
culling
Discarding database objects that are not visible. Usually this refers to discarding objects
located outside the current viewing frustum. This is done by comparing the bounding
volume of these database objects with the six planes that bound the frustum. Objects

007-1680-080

741

Glossary

completely outside may be safely discarded. See also occlusion culling.
data fusion
OpenGL Performer’s ability to read data in a variety of different database formats and
convert it into the internal OpenGL Performer scene database format. Further, the ability
of these different formats to provide special run-time behavior using callback functions
or node subclassing and to have these different data formats all active in their native
modes simultaneously.
database paging
Loading databases from disk or network into memory for traversal during real-time
simulation. Database paging is implicit in large area simulations due to the huge
database sizes inherent in any high-resolution earth database. A frequent component of
database paging is texture paging, in which new textures are downloaded to the graphics
system at the same time new geometry is loaded from disk.
debug libraries
OpenGL Performer libraries compiled with debugging symbols left in are known as
debug libraries. These libraries provide greater and more accurate stack trace
information when examining core dumps, such as during application development.
decal geometry
Objects that appear “above” other objects in pfLayer geometry. In a runways-and-stripes
example, the stripes would be the decal geometry. There can be multiple layers of decals
with successively higher visual priorities. See layer geometry.
depth complexity
The “pixel-rendering load” of a frame, which is defined as the total number of pixels
written divided by the number of pixels in the image. For example, an image of two
full-screen polygons would have a depth complexity of 2. It is often observed that
different types of simulation images have predictable depth complexities with values
ranging from a low of 2.5 for high altitude flight simulation to 4 or more for
ground-based simulations. These figures can serve as a guide when configuring
hardware and estimating frame rates for visual simulation systems. OpenGL Performer
fill statistics provide detailed accounting and real-time visualization of depth complexity,
as seen in Perfly.

742

007-1680-080

Glossary

displace decaling
An implementation method for decal geometry that uses a Z-displacement to render
coplanar geometry. The actual displacement used is a combination of a fixed offset and a
range-based scaled offset, which are combined to produce the effective offset.
display list
A list into which graphics commands are placed for efficient traversal. Both OpenGL
Performer and the underlying OpenGL graphics library have their own display list
structures.
DLL
See dynamic link library.
draw mask
A bitmask specified for both pfNodes and pfChannels, which together selects a subset of
the scene graph for rendering. The node and channel draw masks are logically ANDed
together during the CULL traversal, which prunes the node if the result is zero. Draw
masks may be used to “categorize” the scene graph, where each bit represents a
particular characteristic. Each node contains these masks, binary values whose bits serve
as flags to indicate if the node and its children are considered drawable, intersectable,
selectable, and so on. Most of these bits are available for application use.
drop
Refers to frame processing. When in locked or fixed phase and a processing stage takes
too long, the frame is dropped and not rendered. Dropped frames are a sure sign of
system overload.
DSO
See dynamic shared object.
dynamic
Something that is updated automatically when one of its attributes or children in a scene
graph changes. Often refers to the update of hierarchical bounding volumes in the scene
graph.
dynamic link library (DLL)
The Microsoft Windows equivalent of a DSO.

007-1680-080

743

Glossary

dynamic shared object (DSO)
A library that is not copied into the final application executable file but is instead loaded
dynamically (that is, when the application is launched). Since DSOs are shared, only one
copy of a given DSO is loaded into memory at a time, no matter how many applications
are using it. DSOs also provide the dynamic binding mechanism used by the OpenGL
Performer database loaders.
Euler angles
A set of three angles used to represent a rotation.
See heading, pitch, and roll.
fade count
The application sets a fade count to control the number of frames over which a new DTR
level is faded in.
fixed frame rate
Rendering images at a consistent chosen frame rate. Fixed frame rates are a central theme
of visual simulation and are supported in OpenGL Performer using the
PFPHASE_LOCK and PFPHASE_FLOAT modes. Maintaining a fixed frame rate in
databases of varying complexity is difficult and is the task of OpenGL Performer stress
processing, which changes LOD scales based in measured system load.
flattening
Flattening consists of taking multiple instances of a single object and converting them
into separate objects (deinstancing) and then applying any static transformations defined
by pfSCS nodes to the copied geometry; this action improves performance at a cost in
memory space.
flimmering
The visual artifact associated with improperly drawing coplanar Z-buffered geometry.
For synonyms, consider flitter, flicker, sparkle, twinkle, and vibrate. One way to
understand flimmering is to consider the screen space interpolation of Z-depth values,
wherein a discrete difference of depth must be interpolated across a discrete number of
pixels (or sub-pixels). When two polygons that would be coplanar in an infinite precision
real-number context are considered in this discrete interpolation space, it is clear that the
interpolated depth values will differ when the delta-Z to delta-pixels ratios are relatively
prime. The image that results is essentially a Moirè pattern showing the modular
relationship of the differences in the least significant bits of interpolated depth between
the polygons. The libpr pfDecal() function and libpf pfLayer() node exist to handle

744

007-1680-080

Glossary

the drawing of coplanar geometry without flimmering.
floating phase
The style of frame overload management where the next frame after an overloaded
frame is allowed to start at any vertical retrace boundary rather than being forced to wait
for a specific boundary as in the LOCKED phase.
frame
The term frame is used to mean “image” in most OpenGL Performer contexts. The image
being rendered by the hardware is drawn into a “frame buffer” which is simply an image
memory. This image, when delivered by video signals to a monitor or projector, exists as
one or two video fields. In the one-field case, also known as non-interlaced, each row of
the image is read from the frame buffer and generated as video in sequential order. In the
interlaced method, the first field of display comprises alternate lines, one field for the odd
lines and one field for the even lines. In this mode a frame consists of two fields, as the
norm for NTSC broadcast video. Also, the frame is the unit of work in OpenGL
Performer; the main loop in any Performer application consists of calls to pfFrame().
frame-accurate
In a pipelined multiprocessing model, at any particular time the different stages of the
pipeline are working on different frames. Data in the pipeline is called frame-accurate
when a change made to the data in a particular frame is not visible in downstream stages
of the pipeline until those stages begin processing that frame. Processing of libpf
objects are frame-accurate because multiple copies of data are retained for the different
pipeline stages.
free-running
The unconstrained phase relationship of image generation where frame rendering is
initiated as soon as the previous frame is complete without consideration of a minimum
or maximum frame rate.
frustum
A truncated pyramid—two parallel rectangular faces, one smaller than the other, and
four trapezoidal faces that connect the edges of one rectangular face with the
corresponding edges of the other rectangular face.
gaze vector
The +Y axis from the eyepoint—informally, the direction the eye is facing.

007-1680-080

745

Glossary

graph
A network of nodes connected by arcs. An OpenGL Performer scene graph is so termed
due to its having this form. In particular, an OpenGL Performer scene graph must be an
acyclic graph. See also scene graph.
graphics context
The set of modes and other attributes maintained by OpenGL in both system software
and the hardware graphics pipeline that defines how subsequent geometry is to be
rendered. It is this information which must be saved and restored when drawing occurs
in multiple windows on a single graphics pipeline.
graphics state elements
Individual libpr state components, such as material color, line stipple pattern, point
size, current texture definition, and the other elements that comprise the graphics
context.
heading
In the context of X-axis to the right, Y-axis forward, and Z-axis up, the heading is rotation
about the Z-axis. This is the disturbing rotation that pivots your car clockwise or
counterclockwise during a skid. Heading is also known as yaw, but OpenGL Performer
uses the term heading to keep the H, P, and R abbreviations distinct from X, Y, and Z.
Also see Euler angles.
heap
The process heap is the normal area from which memory is allocated by malloc() when
more memory is required; sbrk() is automatically called to increase the process virtual
memory. Also see arena.
identity matrix
A square matrix with ones down the main diagonal and zeroes everywhere else. This
matrix is the multiplicative identity in matrix multiplication.
immediate-mode rendering
Immediate-mode rendering operations are those that immediately issue rendering
commands and transfer data directly to the graphics hardware rather than compiling
commands and data into data structures such as display lists. See compiled mode.

746

007-1680-080

Glossary

instancing
An object in the scene is called instanced if there is more than one path through the scene
graph that reaches it. Instancing is most commonly used to place the same model in more
than one location by instancing it under more than one pfDCS transformation node.
intersection pipeline
Like the rendering pipeline, OpenGL Performer supports a two-stage multiprocessing
pipeline between the APP and ISECT processes. See also rendering pipeline.
latency
The amount of time between an input and the response to that input. For example,
rendering latency is usually defined as the time from which the eyepoint is set until the
display devices scan out the last pixel of the first field corresponding to that eyepoint.
latency-critical
Operations that must be performed during the current frame and that will reliably finish
quickly. An example of this would be reading the current position of a head-tracking
device from shared memory.
layer geometry
Objects that appear “above” other objects in pfLayer geometry. In a runways-and-stripes
example, the stripes would be the decal geometry. There can be multiple layers of decals
with successively higher visual priorities. See also base geometry.
level of detail (LOD)
The idea of representing a single object, such as a house, with several different geometric
models (a cube, a simple house, and a detailed house, for example) that are designed for
display at different distances. The models and ranges are designed such that the viewer
is unaware of the substitutions being made. This is possible because distant objects
appear smaller and thus can be rendered with less detail. The OpenGL Performer pfLOD
node and the associated pfLODState implement this scheme.
libpf
One of OpenGL Performer’s two core libraries. libpf manages multiprocessing and
scene graph traversals. Built on top of libpr. Multiple copies of libpf objects are
automatically maintained so that the APP, CULL, and ISECT stages of the processing
pipeline do not collide.

007-1680-080

747

Glossary

libpfdu
OpenGL Performer’s database utilities library. It is layered on top of libpf and libpr
and includes functions for building and optimizing geometry before putting it into a
scene graph.
libpfutil
OpenGL Performer’s general utility library, which is distributed in source form for both
usage and information.
libpr
One of OpenGL Performer’s two core libraries. libpr manages graphics state and
rendering, while also providing a number of math and shared memory utility functions.
Provides the foundation for libpf. All processes share the same copy of libpr objects.
libpr classes
The low-level structured data types of libpr. These objects—with the exception of
pfCycleBuffers—lack the special multibuffered, multiprocess data-exclusion support
that libpf objects provide.
light point
A point of light such as a star or a runway light. Accurate display of light points requires
that they attenuate and fog differently than other geometry (see punch through). In flight
simulation, light points often have additional parameters concerning angular
distributions of illumination.
load
The processing burden of rendering a frame. This includes both processing performed on
the host CPU and in the graphics subsystem. It is the maximum of these times (sum in
single process mode) that is used to compute the system stress level for adjusting
pfLODState values.
locked phase
A style of frame-overload processing where drawing may only begin on specific vertical
retraces, namely those that are an integer multiple of the basic frame rate.
morph attribute
One of the collections of arrays of floating point data used in the pfMorph node’s linear
combination processing. This process multiplies each element of each source array by a
changeable weight value for that source array and sums the result of these products to

748

007-1680-080

Glossary

produce the destination array.
morphing
The mathematical manipulation of pfGeoSet data (positions, normals, colors, texture
coordinates) to cause a shape-shifting behavior. This is very useful for animated
characters, continuous terrain level of detail, smooth object level of detail, and a number
of advanced applications. In OpenGL Performer, morphing is provided by the pfMorph
node.
multiple inheritance
Deriving a class from more than one other class. This is in contrast to single inheritance
in which a type hierarchy is a tree. OpenGL Performer does not use multiple inheritance.
multithreaded
In the context of OpenGL Performer culling, multithreading is an option for increased
parallelism when multiple pfChannels exist in a single OpenGL Performer rendering
pipeline. In this case, multiple cull processes are created to work on culling the channels
of a pfPipe in parallel. For example, a single OpenGL Performer pipeline stage (such as
the CULL) is multithreaded when configured as multiple, concurrent processes. These
“threads” are not arranged in pipeline fashion but work in parallel on the same frame.
mutual exclusion
Controlling access to a data structure so that two or more threads in a multiprocessing
application cannot simultaneously access a data structure. Mutual exclusion is often
required to prevent a partially updated data structure from being accessed while it is in
an invalid state.
node
An OpenGL Performer libpf data object used to represent the structure of a visual
scene. Nodes are either leaf nodes that contain libpr geometry or are internal nodes
derived from pfGroup that control and define part of the scene hierarchy.
nonblocking file access
A method of obtaining data from a file without having to wait for any other processes to
finish using the file. Such accesses involve a two-step transaction in which the
application first indicates the task to be performed and is given a handle. This handle can
later be used to inquire about the status of the file action: it is in progress, it has
completed, or there has been an error.

007-1680-080

749

Glossary

non-degrading priorities
Process priorities are used by the operating system to decide when and for how long
processes should run. A non-degrading priority specifies that the process scheduling
should not take into account how long the process has been running when deciding
whether to let another process run. The use of non-degrading priorities is important for
real-time performance.
occlusion culling
The discarding of objects that are not visible because they are occluded by other closer
objects in the scene, for example, a city behind a mountain. See also culling.
opera lighting
The generic term for a powerful carbon-arc lamp producing an intense light such as that
invented by John H. Kliegl and Anton T. Kliegl for use in public-staged events and
cinematographic undertakings that is often mounted within a dual-gimballed
exoskeletal framework to afford the lamp sufficient freedom of orientation that the
projected beam can be made to track and highlight performers as they move across a
stage. The temperature of the thermal plasma that develops between the carbon
electrodes of such arc lamps can be determined by spectroscopic investigation of its
dissociated condition and has been found to be between 20,000˚C and 50,000˚C. The
term can also refer to a stage-lighting technique that projects an image of a background
scene onto the stage or screen. Accurate visual simulation of both of these light types (as
well as common vehicle headlights, airplane landing lights, and searchlights) is provided
by the projected texture capability of the pfLightSource node.
overload
A condition where the time taken to process a frame is longer than the desired frame rate
allows. This causes the goal of a fixed-frame rate to be unattainable and, thus, is an
undesired situation.
overrun
A synonym for overload in the context of fixed frame rate rendering.
pair-wise morphing
The geometric blending of two topologically equivalent objects. Usually this is done by
specifying weights for each object, for example, 90% of object A plus 10% of object B.
Each vertex in the resulting object is a linear interpolation between the vertices in the
original object. See morphing.

750

007-1680-080

Glossary

parent
The OpenGL Performer node directly above a given node is known as the parent node.
passthrough data
Data that is passed down the steps in the rendering pipeline until it reaches a callback.
Such data provides the mechanism whereby an application can communicate
information between the APP, CULL, and DRAW stages in a pipelined manner without
code changes in single-CPU and multiprocessing applications.
path
A series of nodes from a scene graph’s root down to a specific node defines a path to that
node. When there are multiple paths to a node (thus, the scene graph is really a graph
rather than a tree), this path can be important when interpreting an intersection or
picking request. For example, if a car model uses instancing for the tires, just knowing
that a tire is picked is not sufficient for further processing.
Perfly
The application distributed with OpenGL Performer that serves as a demonstration
program. For IRIX and Linux, it is installed in /usr/sbin as well as a programming
example found in /usr/share/Performer/src/sample/apps/C and
/usr/share/Performer/src/sample/apps/C++ for the C and C++ versions,
respectively. For Microsoft Windows, it is installed in %PFROOT%/Bin as well as a
programming example found in %PFROOT%/Src/sample/apps/C and
%PFROOT%/Src/sample/apps/C++ for the C and C++ versions, respectively.
phase
An application’s synchronization mode—defining how the system behaves if the
processing and drawing time for a given frame extends past the time allotted for a frame.
See also locked phase and floating phase.
pipe
Used to refer to both an OpenGL Performer software rendering pipeline and to a
graphics hardware rendering pipeline, such as a RealityEngine. See rendering pipeline.
pitch
In the context of X-axis to the right, Y-axis forward, and Z-axis up, the pitch is rotation
about the X-axis. This is the rotation that would raise or lower the nose of an aircraft. Also
see Euler angles.

007-1680-080

751

Glossary

popping
The term for the highly noticeable instantaneous switch from one level of detail to the
next when morph or blend transitions are not used. This problem is distracting and
should be eliminated in high-quality simulation applications.
process callbacks
The mechanism through which a developer takes control of processing activities in the
various OpenGL Performer traversals and major processing stages: the application
traversal, the cull traversal, the draw traversal, and the intersection traversal all provide
a mechanism for registered process callbacks. These are user functions that are invoked
at the beginning of the indicated processing stage and in the process handling the
traversal.
projective texturing
A texture technique that allows texture images to be projected onto polygons in the same
manner as a slide or movie projector would exhibit keystone distortion when images are
cast non-obliquely onto a wall or screen. This effect is perfect for projected headlights and
similar lighting effects.
prune
To eliminate a node from further consideration during culling.
punch through
Decreasing the rate at which intensely luminous objects such as light points are
attenuated as a function of distance. Normal fogging is inappropriate for such objects
because up close they are actually much brighter than can be rendered given the dynamic
range of the framebuffer and raster display devices.
reference counting
The counter within each pfObject and pfMemory object that keeps track of how many
other data structures are referencing the particular instance. The primary purpose is to
indicate when an object may be safely deleted because it is no longer referenced.
rendering pipeline
An OpenGL Performer rendering pipeline, represented in an application by a pfPipe.
Typically a rendering pipeline has three stages: APP, CULL, and DRAW. These stages
may be handled in separate processes or combined into one or two processes.

752

007-1680-080

Glossary

right-hand rule
Derived from a simple visual example for the direction of positive rotation about an axis,
the right-hand rule states that the curled fingers of the right hand indicate the direction
of positive rotation when the right hand is placed about the desired axis with the thumb
pointing in the positive direction. The positive angle is the one that rotates the primary
axes toward each other. For example, a positive rotation (counterclockwise) about the
X-axis takes the positive Y-axis into the position previously occupied by the positive
Z-axis.
roll
In the context of X-axis to the right, Y-axis forward, and Z-axis up, the roll is rotation
about the Y-axis. This is the rotation that would raise and lower the wings of an aircraft,
leading to a turn. Also see Euler angles.
scene
A collection of geometry to be rendered into a pfChannel.
scene complexity
The complexity of the scene for rendering purposes, in particular the amount of
geometry, transformations, and graphics state changes in the scene.
scene graph
A hierarchical assembly of OpenGL Performer nodes linked by explicit attachment arcs
that constitutes a virtual world definition for traversal and subsequent display.
search path
A list of directory names given to OpenGL Performer to specify where to look for data
files that are not specified as full path names.
sense
An indication of whether a positive angle is interpreted as representing a clockwise (CW)
or counterclockwise (CCW) rotation with respect to an axis. All CCW rotations in
OpenGL Performer are specified by positive (+) angles and negative angles represent
CW rotations.
shadow map
A special texture map created by rendering a scene from the view of a light source and
then recording the depth at each pixel. This Z-map is then used with projective texturing
in a second pass to implement cast shadows. The entire process is automated by the

007-1680-080

753

Glossary

OpenGL Performer pfLightSource node.
share groups
The attributes that a slave share mask can track are divided into groups called share
groups.
share mask
The share mask associates master and/or slave cliptextures.
shared instancing
The simplest form of instancing whereby two or more parent nodes share the same node
as a child. In this situation, any change made to the child will be seen in each instance of
that node. Also see cloned instancing.
shininess
The coefficient of specular reflectivity assigned to a pfMaterial that governs the
appearance of highlights on geometry to which it is bound.
siblings
The name given to nodes that have the same parent in a scene graph.
skip
Refers to frame processing. See drop.
sorting
The grouping together of geometry with similar graphics state for more efficient
rendering with fewer graphics state changes. OpenGL Performer sorts during scene
graph traversal.
spacing
The relative motion required to move the starting point for subsequent pfFont rendering
after drawing a particular character pfGeoSet in a pfFont. This motion is a pfVec3 to
allow arbitrary escapement for character sets that use vertical rather than horizontal text
layouts. Note that for vertically oriented fonts, the origin should be such that motion by
the spacing value crosses the character; in other words, the origin should be on the left
for left-to-right fonts, at the top for top-to-bottom fonts, on the bottom for bottom-to-top
fonts, and on the right for right-to-left fonts.

754

007-1680-080

Glossary

spatial organization
The grouping together of geometric objects that are spatially close to each other in the
scene graph. For optimal culling performance, the scene should be organized spatially.
sprite
A transformation that rotates a piece of geometry, usually textured, so that it always faces
the eyepoint.
stage
This is a section of the OpenGL Performer software rendering pipeline, either
application, culling, or drawing. Sometimes it is used to refer to either of the two
non-pipeline tasks of intersection and asynchronous database processing.
state
Refers to attributes used to render an object that are managed during traversal. State
commonly falls into two areas: traversal state that affects which portions of the scene
graph are traversed, and graphics state that affects how something is rendered. Graphics
state includes the current transformation, the graphics modes managed by pfGeoStates,
and other states such as stenciling.
stencil decaling
An implementation method for pfLayer nodes that uses an extra bit per pixel in the frame
buffer to record the Z-buffer pass or fail status of the base geometry. This bit is then used
as a visibility determination (rather than the Z-buffer test) for each of the layers, which
are rendered in bottom (lowest visual priority) to top (highest visual priority) order.
Z-buffer updating is disabled during the stencil rendering operation and is restored
when the pfLayer node has been completely rendered. Stencil-bit processing is the
highest quality mode of pfLayer operation.
stress
OpenGL Performer stress processing is the closed-loop feedback mechanism that
monitors cull and draw times to determine how pfLODState range scale factors should
be adjusted to compensate for system load in order to maintain a chosen frame rate.
subgraph
A connected subset of a scene graph; usually, the set consisting of all descendents of a
particular node.

007-1680-080

755

Glossary

texel
Short for “texture element”—a pixel of a texture.
texture mapping
Displaying a texture as though it were the surface of a given polygon.
tile
A section of a spatially subdivided database or a rectangular subregion of a larger texture
image.
transformation
Homogeneous 4x4 matrices that define 3D transformations—some combination of
scaling, rotation, and translation.
transition distance
The distance at which one level-of-detail model is switched for another. When fading or
morphing between levels-of-detail, the distance at which 50% of each model is rendered.
See level of detail.
traversals
One of OpenGL Performer’s pre-order visitations of a hierarchical scene graph.
Traversals for application, culling, and intersection processing are internal to libpf and
user-written traversals are supported by the pfuTraverser tools.
traversing
See traversals.
trigger routine
A routine that initiates a traversal or the invocation of a callback in another process.
pfCull() triggers the cull traversal. pfFrame() triggers processing for the current frame.
up vector
The +Z axis of the eyepoint; it defines the display’s “up” direction. Must be
perpendicular to the gaze vector.

756

007-1680-080

Glossary

view volume visualization
The display of the viewing frustum for a particular channel, usually done by rendering
a wireframe version of the frustum with a different eyepoint or field-of-view. See cull
volume visualization.
viewing frustum
The frustum containing the portion of the scene database visible from the current
eyepoint.
viewpoint
The location of the camera or eye used to render the scene.
viewport
The portion of the framebuffer used for rendering. Each pfChannel has a viewport in the
framebuffer of its corresponding pfPipeWindow.
visual
A construct that the X Window System uses to identify framebuffer configurations.
widget
A manipulable or decorative element of a graphical user interface. Much of the
programming for GUI elements is associated with defining the reaction of widgets to
user-mouse and keyboard events.
window manager
A special X Window System client that handles icons, window placement, and window
borders and titles.

007-1680-080

757

Index

Numbers
2D tiles, 365
3DS format. See formats

A
accessing GL, 313
acreate(), 496, 726, 737
activation of traversals, 86
active database
billboards, 75
active scene graph. See application traversal
active surface definition, 147, 535
addQueryArray, 559
addQueryGeoSets, 559
affine transformations, 615
alias, definition, 737
align geometry, 562
allocating memory. See memory
alpha function, 315
animation, 67, 532
using quaternions for, 617
anisotropic filtering, 326
antialiasing, 318, 367
APP, 21
application areas
rapid rendering, xxxix
simulation-based design, xxxix

007-1680-080

virtual reality, xxxix
virtual sets, xxxix
visual simulation, xxxix
application buffer, 164
defined, 737
application traversal, 88
ArchVision RPC database format, 273
arenas, 496
defined, 737
See also shared memory
arithmetic, precision of, 672
ASD, 147, 535, 543
and pfEngine, 564
cliptexture, 561
flow chart, 540
paging, 566
simple example, 541
vertices, 544
aspect ratio matching, 28
assembly mock-up, xxxix
assignment operators, 736
asynchronous database paging, definition, 737
asynchronous database processing, 163
asynchronous deletion, 164
asynchronous I/O, 502
atmospheric effects
enabling, 185
attribute binding, definition, 737
attribute data structure, 554
attributes, 548, 554

759

Index

bindings, 301, 668
flat-shaded, 302
global, 555
overview, 299
traversals, 86
AutoCAD, 250
automatic type casting, 7
average statistics, 653
See also statistics
axes, default, 30
axially aligned boxes, 621

B
backface culling, 160, 173, 174
base classes, 6
base geometry, 316
definition, 738
basic-block counting, 673
behaviors, 88
bezel, definition, 738
billboards, 75, 671
defined, 738
implementation using sprites, 335
BIN format. See formats
binary operators, 736
binning, definition, 738
bins, 97
definition, 738
blended transparency, 315
blur, 424
blur margin, 427, 738
bottlenecks, 172, 660
fill, 662
host, 660
transform, 661

760

bounding volumes
defined, 738
example code, 59
See also volumes
buffer scope, 164
defined, 739
BYU format. See formats

C
C++, See OpenGL Performer C++ API
caching
intersections, 667
state changes, 657
callbacks
culling, 100, 112-115
customized culling, 90
discriminators for intersections, 629
draw, 112-115
function, 112
node, 112
post-cull, 113
post-draw, 113
pre-cull, 113
pre-draw, 113
process, 115
calligraphic light point, 587
calligraphic lights, number of, 590
calligraphic vs. raster displays, 588
calligraphic, color correction, 597
calligraphic, simulating, 602
CASEVision, 672
channels, 386
channel share group, 40
definition, 739
configuring, 26
creating, 26
definition, 739

007-1680-080

Index

multiple, rendering, 35
share mask, definition, 739
children, of a node, definition, 739
class hierarchy, definition, 739
class inheritance, 7
definition, 740
classes
libpf
pfBillboard, 51, 75, 721
pfBuffer, 163, 722
pfChannel, 21, 26, 359, 722
pfCullProgram, 103
pfDCS, 51, 62, 64, 88, 721
pfDoubleDCS, 51, 721
pfDoubleFCS, 51, 721
pfDoubleSCS, 51, 721
pfEarthSky, 27, 181, 722
pfFrameStats, 635
pfGeode, 51, 71, 721
pfGroup, 51, 722
pfIBRnode, 204, 273
pfIBRtexture, 206, 273
pfLayer, 51, 70, 722
pfLightPoint, 722
pfLightSource, 51, 722
pfLOD, 51, 70, 722
pfLODState, 722
pfMPClipTexture, 386, 722
pfNode, 49, 51, 52, 722
pfPartition, 51, 78, 722
pfPath, 723
pfPipe, 21, 23, 355, 722
pfPipeVideoChannel, 723
pfPipeWindow, 21, 359, 469, 723
pfRotorWash, 723
pfScene, 19, 51, 61, 722
pfSCS, 51, 62, 64, 88, 722
pfSequence, 51, 67, 722
pfShader, 723
pfShaderManager, 723

007-1680-080

pfShadow, 199
pfSwitch, 51, 67, 722
pfText, 51, 722
pfTraverser, 723
pfVolFog, 186, 723
libpfdu
pfdBuilder, 233
pfdGeom, 237
pfdPrim, 238
libpfv
pfvDisplayMngr, 44
pfvInteractor, 712
pfvPicker, 712
pfvSelector, 712
pfvViewer, 44, 700
libpr
pfBox, 621, 724
pfClipTexture, 386, 723
pfColortable, 723
pfCompositor, 367
pfCycleBuffer, 499, 724
pfCycleMemory, 499, 724
pfCylinder, 621, 724
pfDataPool, 498, 724
pfDispList, 658, 723
pfFBState, 723
pfFile, 724
pfFog, 723
pfFont, 305, 723
pfFrustum, 724
pfGeoSet, 291, 657, 723
pfGeoState, 723
pfHighlight, 723
pfHit, 628, 723
pfLight, 724
pfLightModel, 724
pfList, 725
pfLoadBalance, 370
pfLPointState, 724
pfMaterial, 724
pfMatrix, 613, 725

761

Index

pfMatStack, 619, 725
pfMemory, 725
pfObject, 725
pfPassList, 723
pfPlane, 622
pfPolytope, 724
pfQuat, 617, 725
pfQueue, 386
pfSeg, 626, 724
pfSegSet, 120, 724
pfSphere, 621, 724
pfSprite, 335, 724
pfState, 724
pfStats, 635, 653, 725
pfString, 307, 724
pfTexEnv, 724
pfTexGen, 724
pfTexture, 396, 724
pfType, 725
pfVec2, 611, 725
pfVec3, 611, 725
pfVec4, 611, 725
pfWindow, 725
clip center, 374, 382
clip center node, 418
clip region, 374
clip size, 374
clip_size, 410
clipped level, 377
clipped, definition, 740
clipping planes, definition, 740
cliptexture, 373, 393, 740
and ASD, 561
center, 384
configuration, 392, 397
inset, 437
invalidating, 428
load control, 423
loaders, 450

762

manipulating, 423
multipipe applications, 445
multiprocessing, 415
preprocessing, 389
read queue, 428
sample code, 448
slave, 446
slave and master, 420, 445
test and demo programs, 449
utility code, 451
virtual, 382, 429, 446
with multiple pipes, 420
clocks
high-resolution, 494
cloned instancing, 57
definition, 740
clones, 362
cloning, definition, 740
close(), 502
closed loop control system, 152
code examples. See example code
color correction, 597
compiled mode, 294
definition, 741
complex pixels, definition, 741
compositors, 351, 365, 367
computer aided design, xxxix
configuration
cliptexture, 392, 397
cliptexture files, 399
cliptexture utilities, 397
default tile, 396
files
cliptextures, 399
MPK, 44
XML, 702, 716
image cache, 395, 398, 401, 402, 403
image cache levels, 394
image cache proto tile, 395

007-1680-080

Index

image tile, 396
load time, 393
optional image cache, 414
pfChannel, 26
pfFrustum, 28
pfPipe, 23
pfPipeWindow, 469
pfScene, 27
pfTexture, 396
viewpoint, 30
viewport, 27
configuration file
creating, 400
containment, frustum, 93
conventions
typographical, xlii
coordinate systems, 30
dynamic. See pfDCS nodes
static. See pfSCS nodes
coplanar geometry, 70, 316
copying pfObjects, 15
core dump
from aggregate pfObject allocation, 725
from mixing malloc() and pfFree(), 496
from mixing pfMalloc() and free(), 496
from static pfObject allocation, 725
from unshared pfObject allocation, 725
Coryphaeus
DWB format, 222
cost tables, 426, 741
counter, video, 495
counting, basic-block, 673
CPU statistics, 641
critically damped, definition, 741
CULL, 21
backface culling, 160, 173, 174
sidekick processes, 159, 173-179
cull program, 103, 105

007-1680-080

cull volume visualization, definition, 741
culling
callbacks, 90
definition, 741
efficient, 94
multithreading, 159
See also traversals
traversal, 90
cull-overlap-draw multiprocessing model, 157
cumulative statistics, 653
See also statistics
current statistics, 653
See also statistics
cut-ins, 365
cycle buffers, 171, 499
cylinders
as bounding volumes, 621
bounding, 667

D
data fusion
defined, 742
data structures, 547
database formats
RPC, 273
SGF, 275
database loaders, 730
database paging, 94, 163
definition, 742
databases
importing, 221
optimization, 669
organization, 88, 94
traversals, 85-127
datapools. See pfDataPool data structures
dbx
See debugging

763

Index

DCS. See pfDCS nodes
debug libraries, definition, 742
debugging
dbx, 676
guidelines, 675
ogldebug, 674
shared memory and, 675
decal geometry, definition, 742
decals, 559
See also coplanar geometry
decomposition
hardware decomposition, 365
spatial, 352
temporal, 352
decomposition modes, 367
default tile
configuration, 396
deleting objects, 11
depth complexity, definition, 742
detail texture, 671
device, streaming, 407
disabling graphics modes, 318
discriminator callbacks
for intersections, 629
displace decaling, 316
defined, 743
display list, 337, 658
definition, 743
display list mode, 294
display, raster vs. calligraphic, 588
display, stereo, 38
displaying statistics. See statistics
DLL (dynamic link library) files, 223
dlopen(), 224, 228
dlsym(), 224, 228
double-precision arithmetic, 672
double-precision matrices, 64, 67

764

download time
cliptexture, 425
DPLEX (Video Digital Multiplexer), 351
DRAW, 21
DRaW Computing Associates, 287
draw mask, 111
draw mask, definition, 743
draw traversals. See traversals
drop, definition, 743
DSO (dynamic shared object) files, 223
DTR, 388, 424
DVR, 148
DWB format. See formats
DXF format. See formats
dynamic coordinate systems. See pfDCS nodes
dynamic link library (DLL) files, 223
dynamic shared object (DSO) files, 223, 744
Dynamic Texture Resolution, 424
dynamic video resolution, 148
dynamic, definition, 743

E
earth/sky model, 27
effective levels, 383
effects, atmospheric, enabling, 185
elastomeric propulsion system, 31
enabling
atmospheric effects, 185
fog, 185
graphics modes, 318
statistics classes, 647
engine, and ASD, 564
enterView(), 712
enterWorld(), 712

007-1680-080

Index

environment variables
DISPLAY, 464
LD_LIBRARY_PATH, 225, 673
PFHOME, 225
PFLD_LIBRARY_PATH, 225
PFNFYLEVEL, 503
PFPATH, 504, 505
PFTMPDIR, 497
PROFDIR, 674
environmental model, 27
error handling
floating-point operations, 676
notification levels, 502
Euler angles, 30
defined, 744
evaluation function, 546
default, 556
overriding, 557
timing, 561
EventView performance analyzer, 682-697
example code
bounding volume, 59
C++
invalid creation of objects, 726
valid creation of objects, 726
calligraphic lights, 603
converting to pfFlux, 521
convex polyhedrons, 627
creating cloned instances, 59
cull-process callbacks, 116
database loaders, 730
default stress function, 154
fill statistics, 644
frame control, 135
graphics state, 338
hierarchy construction, 55
highlighting, 334
hyperpipes, 355, 356
intersection routines, 630
libpf, 731

007-1680-080

libpfutil, 664, 666
light point process, 586
LOD, 143
matrices and vectors, 616
MIPmap, 321
Motif, 481
passthrough data, 118
Perfly, 751
pfASD, 564
pfBillboard, 76
pfCalligraphic, 599
pfChannel, 32, 39, 42
pfCopy(), 15
pfDecal, 317
pfDelete(), 13
pfEarthSky(), 182
pfFont, 306
pfGeoSet, 518
pfGeoSets, 72
pfGeoStates, 341
pfLayer, 71
pfNode, 114
pfOverride, 339
pfPartition, 79
pfPipe, 363
pfPipe use, 25
pfPipeWindow, 474, 476, 479, 480
pfPipeWindows, 362
pfScene, 53
pfSequence, 68
pfString, 307
pfStrings, 73
pfSwitch, 68
pfTexture, 321
pfWindow, 454, 465, 467
raster callback, 581
scene graph traverser, 17
static member data, 734
statistics, 645, 654
transformation inheritance, 80
user data, 11

765

Index

X input, 666
exceptions, floating-point, 676
exec(), 675
exitView(), 712
exitWorld(), 712
ext_format, 404, 410
extending bounding volumes, 623
extensibility
callback functions, 732
subclassing, 729
user data, 10

F
face culling, 317
fade count, 426, 744
field of view, 28
field, video, 637
files
database formats, 241
loading. See databases
fill statistics, 644
See also statistics
filter
stress filter, 151
fixed frame rates, 129
defined, 744
flat-shaded line strip, 298
flat-shaded primitives, 294
flatten, definition, 744
FLIGHT format. See formats
flimmering, 316, 744
floating phase, definition, 745
floating-point exceptions, 664, 676
flux sync groups, 515
fog, 186

766

atmospheric effects, 183
configuring, 332
data structures, 184, 332
enabling, 185
performance cost, 661
fork(), 168, 675, 733
formats
3DS, 244
BIN, 244
BYU, 248
DWB, 249
DXF, 250
FLIGHT, 252
GDS, 254
GFO, 254
IM, 256
IRTP, 257
LSA, 259
LSB, 259
MEDIT, 263
NFF, 263
OBJ, 265
Open Inventor, 258
PHD, 268
POLY, 246
PTU, 270
SGO, 276
SPF, 280
SPONGE, 280
STAR, 281
STL, 282
SV, 283
TRI, 287
UNC, 287
VRML, 258
FOV. See field of view
frame accurate, definition, 745
frame rate, 148
frame(), 711

007-1680-080

Index

frames
definition, 745
management, 129
overrun, 132
synchronization, 132
free(), 496
free-store management, 11
frustum, 28
as camera. See channel
as culling volume, 622
definition of, 745
FSAA (See full-scene antialiasing.)
full-scene antialiasing (FSAA), 366
function callbacks, 112
functions
naming, 2
See also routines and OpenGL functions.

graphics
attributes, 311
load. See load management
modes, 311, 313
pipelines. See pipelines
state, 311
state elements, definition, 746
statistics, 643
See also statistics
values, 311, 318
graphics context, definition, 746
graphics libraries
database sorting, 665
input handling, 666
objects, 671
OpenGL, xxxix
See also OpenGL
graphics pipe, 351
grout, digital, 147

G
gaze vector, definition, 745
GDS format. See formats
genlock, 495
geometry
coplanar. See coplanar geometry
nodes, 71
rotating, 75, 671
volumes. See volumes
getenv(), 505
getting started, xxxix
GFO format. See formats
global attribute, 555
global state, 340
GLXFBConfigSGIX, 459
graph
defined, 746
stage timing. See stage timing graph

007-1680-080

H
half-spaces, 622
handleEvent(), 709, 711
handling flimmering, 70
hardware composition schemes, 365
header file, 2, 721
header_offset, 413
heading, 30
defined, 746
heap, 737
defined, 746
help
accessing the mailing list, xliii
C++ argument passing, 728
channel groups, 40
channels, 26
clearing a channel, 118

767

Index

database formats, 241
database paging, 163
default shared arena size, 498
display lists, 294
drawing a background, xli, 181
drawing text, 73
flimmering, 744
frame rates, 129
geometry specification, 291
graphics attributes, 311
inheriting transformations, 80
instancing, 56
interfacing C and C++ code, 729
level of detail, 136
morphing, 524
multiple pipelines, 22
multiprocess configuration, 23
node callback functions, 112
overview of chapter contents, xl
performance tuning, 655
pipes, 21
scene graph structure, 94
scene graphs, 87
shared memory, 495
traversals, 85
understanding process models, 160
understanding statistics, 636
view specification, 31
viewports, 27
where to start, xxxix
windows, 469
writing a loader, 230
help process callback functions, 115
high-resolution clocks, 494
horizontal stripes, 365
hyperpipe, 351, 352, 359
programming with, 364
hyperpipes
multiple, 354

768

I
I3DM modeler, 283
icache_files, 411, 412
icache_format, 411
icache_params, 411
icache_size, 404
identity matrix, definition, 746
IM format. See formats
image cache, 375, 394, 411
configuration, 395, 398, 400, 401, 403
levels, 394, 411
proto tile, 395
image data, formatting, 391
image tile, 394, 396
image, tiling, 391
image-based rendering, 204
img_format, 404, 410
immediate mode rendering, definition, 746
immediate-mode, 294
include files, 721
index attributes, 303
indexed pfGeoSets, 291
industrial simulation, xxxix
INF (infinite value) exception, 676
InfinitePerformance graphics systems, 365
info-performer, xliii
inheriting
attributes, 49
classes, 7
state, 87
initializing
C++ virtual functions, 733
pfType system, 733
inline, 611
in-lining math functions, 672
input handling, 666

007-1680-080

Index

inset, 385
adding to cliptexture, 438
and DTR, 438
boundary, 439
building, 439
cliptexture, 437
multiple, 441
inset views, 38
instancing, 56
cloned, 57
definition, 747
shared, 56
int_format, 404, 410
Interest Area, 566
internal API, 720
interpolation, MIPmap, 671
intersections
caching, 667
masks, 122, 629
performance, 667
pipeline, definition, 747
See also discriminator callbacks
tests
geometry sets, 628
planes, 628
point-volume, 624
segments, 627
segment-volume, 627
triangles, 628
volume-volume, 624
traversals. See traversals
invalid border, 380, 388, 411
I/O
asynchronous, 502
handling, 666
IRIS IM, 481
IRIS Image Vision Library, 271
IRIS Inventor. See Open Inventor
IRIX kernel, 675

007-1680-080

IRTP format. See formats
islDrawAction
setLoadTextureFunc, 349
setLoadTextureFunc(), 349
islShape
draw(), 350
setDrawGeometryFunc(), 349

K
Kaleido, polyhedron generator, 268
kernel, 675
keyframing
using quaternions for, 617

L
latency
controlling, 664
defined, 747
latency-critical
definition, 747
updates, 664
layer geometry, 316
definition, 747
layered fog, 186
level of detail, 390, 535
blended transitions, 145
cannonical channel resolution, 142
cannonical field of view, 142
defined, 747
stress management, 640
switching, 70
use in optimization, 659
libpf, 417

769

Index

cliptextures, 386
defined, 747
display environment, 19
objects, 10
libpfct, 401
libpfdb, 221
cliptextures, 387
libpfdu, 221, 222
cliptextures, 387
defined, 748
libpfim, 401
libpfmpk library, 44
libpfspherepatch, 401
libpfutil, 221
cliptextures, 387
libpfv library, 19, 699
libpfvct, 401
libpr
classes, 720, 748
cliptextures, 386, 393
defined, 748
graphics state, 311
objects, 10
Light Point Board, 590
light points, 108
definition, 748
light shafts, 196
lighting
overview, 330
lights, bright, 587
Lightscape Technologies, 259
line segments, 626
See also pfSegSet data structures
load balancing, 370
load control, cliptextures, 388
load management
level-of-detail scaling, 152-155
statistics, 640

770

load, definition, 748
loaders, 230
loading files. See databases
LoadLibrary(), 228
load-time configuration, 393
local state, 340
locked phase, definition, 748
locks, allocating, 498
LOD (level of detail), 536
managing, 136
range, 539
reduction, 537
See also level of detail
See also load management
user control over evaluation, 147
lookahead cache, 375, 377
LPB, 587, 590
LSA. See formats
LSB. See formats

M
macros, 672
PFADD_SCALED_VEC3, 612
PFADD_VEC3, 612
PFALMOST_EQUAL_MAT, 615
PFALMOST_EQUAL_VEC3, 613
PFCOMBINE_VEC3, 612
PFCONJ_QUAT, 618
PFCOPY_MAT, 614
PFCOPY_VEC3, 612
PFDISTANCE_PT3, 612
PFDIV_QUAT, 618
PFDOT_VEC3, 612
PFEQUAL_MAT, 615
PFEQUAL_VEC3, 613
PFGET_MAT_COL, 614
PFGET_MAT_COLVEC3, 614

007-1680-080

Index

PFGET_MAT_ROW, 614
PFGET_MAT_ROWVEC3, 614
PFLENGTH_QUAT, 618
PFLENGTH_VEC3, 612
PFMAKE_IDENT_MAT, 613
PFMAKE_SCALE_MAT, 613
PFMAKE_TRANS_MAT, 613
PFMATRIX, 729
PFMULT_QUAT, 618
PFNEGATE_VEC3, 612
PFQUAT, 729
PFSCALE_VEC3, 612
PFSET_MAT_COL, 614
PFSET_MAT_COLVEC3, 614
PFSET_MAT_ROW, 614
PFSET_MAT_ROWVEC3, 614
PFSET_VEC3, 612
PFSQR_DISTANCE_PT3, 612
PFSUB_VEC3, 612
PFVEC2, 729
PFVEC3, 729
PFVEC4, 729
mailing list, xliii
makeIBRimages program, 219
makeProxyImages program, 215
malloc(), 675
See also memory, pfMalloc()
masks, intersection, 122, 629
master cliptexture, 420
materials, 331
math routines, 611-633
in-lining, 672
matrices
4 by 4, 613
affine, 615
composition order, 616
double-precision, 64, 67
manipulating, 335
stack functions, 619

007-1680-080

matrix routines
transformations, 613
matrix. See transformation
matrix stack, 619
maxlod, 383
measuring performance, 672
Medit, 222
MEDIT format. See formats
mem region, 376
mem_region_size, 404
member functions, 719
overloaded, 730
memory
allocating, 496, 675
multiprocessing, 168
shared. See shared memory
memory mapping, for shared arena, 497
memory requirements, 441
Menger sponge, 280
mesh, 535, 543
methods (See routines.)
Microsoft Developer Studio, 676
Mies, 255
minification, 671
MIPmap, 374
building, 389
filtering, 326
interpolation functions, 671
level, 375
mode changes, 661
modelers
AutoCAD, 250
Designer’s Workbench, 249
EasyScene, 249
EasyT, 249
I3DM, 283
Imagine, 258

771

Index

Kaleido, 268
Model, 265
ModelGen, 252
MultiGen, 252
morph attribute, definition, 748
morphing, 524
defined, 749
morphing vector, 542, 544
Motif, 481
MPK (See OpenGL Multipipe SDK.)
MPK configuration files, 44
MPKConfig class, 47
multibuffering, 735
Multi-Channel Option, 39
multipass rendering, 103
multiple channels, 32, 39, 40
and ASD, 560
rendering, 35
multiple hardware pipelines, 22
multiple inheritance
avoidance of, 10
definition, 749
multiple pipelines. See pipelines
multiplexers, 351
multiprocess, cliptexture, 415
multiprocessing
display-list generation, forcing, 157
functions, invoking during, 166
memory management, 168
models of, 156
cull-overlap-draw, 157
timing diagrams, 160
order of calls, 160
pipelines, 115
pipelines, multiple, 159
uses for, 155
multisampling, 680
multi-texture support, 345

772

multithreading, 159
defined, 749
mutual exclusion, definition, 749

N
NaN (Not a Number) exception, 676
neighborhood array, 556
NFF format. See formats
node draw mask, 111
nodes
callbacks, 112
defined, 749
pruning, 90
sequences, 67
types, 51
nonblocking access, definition, 749
nonblocking file interface, 502
non-clipped level, 377
non-degrading priorities, definition, 750
notification levels for errors, 502
num_streams, 407
numEffectiveLevels, 431

O
OBJ format. See formats
object creation, 4
object derivation, 7
object type, 17
determining, 17
Objects, 725
occlusion culling, definition, 750
ogldebug, 673, 674
utility, 674

007-1680-080

Index

Onyx RealityEngine. See RealityEngine graphics
opcodes, 105
Open Inventor, 89, 226, 258
loader, C++ implementation, 730
open(), 502
OpenGL, xxxix
functions
glAlphaFunc(), 315
glBlendFunc(), 680, 681
glColorMaterial(), 657, 662, 682
glFinish(), 645
glFog(), 192, 333
glLight(), 330
glMaterial(), 331
glShadeModel(), 661
glStencilOp(), 316
glTexEnv(), 320
glTexGen(), 329
glTexImage2D(), 320
glXCreateContext(), 460
glXQueryHyperpipeNetworkSGIX(), 356
glXSwapBuffers(), 133, 364, 674
OpenGL Multipipe SDK (MPK) product, 44
OpenGL Performer
applications of, xxxix
C API, 719
C++ API, 719
accessor functions, 719
header files, 721
member functions, 719
new, 725
object creation, 725
object deletion, 725
public structs, 720
reserved functions, 720
static class data, 734
subclassing, 729
using both C and C++ API, 728
using the C API with C++, 727

007-1680-080

differences between C and C++
error handling, 502
getting started, xl
introduction, xxxix
mailing list, xliii
type system, 17, 720, 721, 730
OpenGL Performer API, 1
OpenGL Shader, 347
opera lighting
defined, 750
operator
delete, 725
new, 725
optimal pfGeoSet size, 657
optimization
database parameters, 669
organization of databases. See databases
orthogonal transformations, 615
orthonormal transformations, 616, 623
overlay(), 712
overload, definition, 750
overrun, definition, 750
overrun, frame, 132

P
paging
in ASD, 566
multi-resolution, 568
order of in ASD, 568
preprocessing, 567
parameters,virtual cliptexture, 431
parent, of a node, defined, 751
parser, 399
partitions, 78
pass-through data
defined, 751

773

Index

passthrough data, 117, 170
patchy fog, 186
paths
definition, 751
search paths, 504
through scene graph, 100
perfly, 135, 636, 644
definition, 751
sample programs, 225
Perfly application, 209
performance
costs
lighting, 661
multisampling, 680
EventView performance analyzer, 682-697
intersection, 667
measurement, 672
tuning
database structure, 667
graphics pipeline, 660
guidelines, specific, 660
optimizations, built-m, 656
overview, 655
process pipeline, 663
RealityEngine graphics, 680
Performance Co-Pilot, 673
Performer Terrain Utilities, 270
PF_DTR_MEMLOAD, 425
PF_DTR_READSORT, 425
PF_DTR_TEXLOAD, 425
PF_MAX_ANISOTROPY, 326
pfAddMPClipTexture(), 416, 421
pfAddPWinPVChan(), 35, 478, 483
pfAppearanceChanged(), 348
pfAppFrame(), 638
pfApplyDecalPlane(), 316
pfApplyTLOD(), 328
pfApplyTMat(), 322

774

pfASD, 552
and pfEngine, 564
queries, 558
PFASD_COLORS, 555
PFASD_NORMALS, 555
PFASD_TCOORDS, 555
pfASDFace, 545
pfASDLODRange, 556
pfASDVert, 553
pfAttachPWinSwapGroup(), 482
pfAttachWin(), 463
pfAttachWinSwapGroup(), 461, 464
PFB file format, 240
pfBillboard, 75
pfBillboard nodes, 671
pfBindPVChan(), 485
pfBindPWinPVChans, 485
pfBox, 621
pfChannel data structures. See channels
pfChanPixScale, 149
pfChanPWinPVChanIndex(), 35, 151
pfChanTravMode(), 110
pfChoosePWinFBConfig(), 473
pfClipTexture, 373
and ASD, 561
pfCompositorAddChild(), 368
pfCompositorChannelClipped(), 368
pfCompositorChildFrustum(), 370
pfCompositorChildSubdivision(), 370
pfCompositorChildViewport(), 370
pfCompositorLoadBalancer(), 369
pfCompositorMasterPipe(), 370
pfCompositorMode(), 369
pfCompositorNearFar(), 368
pfCompositorOrtho(), 368
pfCompositorPerspective(), 368

007-1680-080

Index

pfCompositorReconfig(), 368
pfCompositorVal(), 369
pfCompositorViewport(), 368
pfCompute(), 158
pfComputeFunc(), 159
pfConfig(), 46, 47, 711
pfconv, 240
pfCull(), 712
pfCullProgram, 103
pfCullProgramAddBinParent(), 109
pfCullProgramAddPgmInstruction(), 105
pfCullProgramAddPgmOpcode(), 105
pfCullProgramIsSubbinOf(), 109
pfCullProgramResetBinParents(), 109
pfCullProgramResetPgm(), 105
pfCullProgramTestPolytope(), 109
pfCylinder, 621
pfDataPool data structures, 498
multiprocessing with, 170
pfdBuilder, 256
pfdConverterAttr_rpc(), 273
pfdConverterMode_rpc(), 273
pfdConverterVal_rpc(), 273
pfDCS nodes, 658
pfDeleteGLHandle(), 323
pfdGetConverterAttr_rpc(), 273
pfdGetConverterMode_rpc(), 273
pfdGetConverterVal_rpc(), 273
pfdInitConverter(), 223
pfDispList, 337
pfDispList data structures, 337
pfdLoadClipTexture, 387
pfdLoadClipTextureState, 387
pfdLoadImageCache, 387
pfdLoadImageCacheState, 387

007-1680-080

pfdLoadNeededDSOs(), 224
pfdLoadNeededDSOs_EXT(), 228
pfDoubleDCS nodes, 64
pfDoubleFCS nodes, 67
pfDoubleSCS nodes, 64
pfdProcessASDTiles, 569
pfDraw(), 712
pfdWriteFile, 569
pfEngine, 521
and ASD, 564
pfEvaluateLOD(), 147
pfFilePath(), 504
pfFilePathv(), 504
pfFlux, 507
pfFog data structures, 184, 332
See also fog
pfFont, 305
pfFrame(), 469, 711
pfFrameStats data structures, 635
See also pfStats data structures
pfGeode, 71
pfGeoSet, 291
and bounding volumes, 620
compilation, 294
connectivity, 296
draw modes, 294
intersection mask, 629
intersections with segments, 628
primitive types, 293
pfGeoSet data structures
adding to pfGeode nodes, 12, 71
pfGeoState data structures
applying, 341
attaching to pfGeoSets, 342
overview, 339
pfGetChanOrigin(), 149
pfGetChanOutputOrigin(), 149

775

Index

pfGetChanOutputSize(), 149
pfGetChanPixScale, 149
pfGetChanPWinPVChanIndex, 151
pfGetChanSize(), 149
pfGetCompositorChannelClipped(), 368
pfGetCompositorChildFrustum(), 370
pfGetCompositorChildSubdivision(), 370
pfGetCompositorChildViewport(), 370
pfGetCompositorFrustType(), 368
pfGetCompositorFrustum(), 368
pfGetCompositorLoadBalancer(), 369
pfGetCompositorMode(), 369
pfGetCompositorNearFar(), 368
pfGetCompositorVal(), 369
pfGetCompositorViewport(), 368
pfGetCurCalligraphic(), 602
pfGetGSetPrimLength(), 292
pfGetIBRnodeProxyTexCoords(), 206
pfGetIBRtextureDefaultTexture(), 205
pfGetLoadBalanceLoad(), 372
pfGetLoadBalanceNumActive(), 372
pfGetLoadBalanceVal(), 372
pfGetLoadBalanceWork(), 372
pfGetMPClipTexture(), 417
pfGetNumMPClipTextures(), 417
pfGetNumScreenPVChans(), 484
pfGetPFChanStressFilter(), 152
pfGetPVChanId, 151
pfGetPVChanInfo(), 485
pfGetPWinNumPVChans(), 484
pfGetPWinPVChanId, 484
pfGetPWinPVChanIndex, 484
PFGS_FLAT_TRIFANS, 302
PFGS_PACKED_ATTRS, 295
pfGSetDecalPlane(), 316

776

pfGSetMultiAttr(), 292, 299
pfGSetOptimize(), 179
pfGSetQuickAttr(), 178
pfGSetQuickMultiAttr(), 179
pfGSetQuickPrimLengths(), 179
pfHit, 628
pfHyperpipe, 359
PFI image format, 240
pfIBRnode class, 204, 273
pfIBRnodeAngles(), 205
pfIBRnodeFlags(), 206
pfIBRnodeIBRtexture(), 205
pfIBRnodeProxyTexCoords(), 206
pfIBRtexture class, 206, 273
pfIBRtextureDirection(), 208
pfIBRtextureFlags(), 207, 208
pfIBRtextureIBRdirections(), 206
pfIBRtextureIBRtextures(), 207
pfIBRtextureLoadIBRtexture(), 207
pfIBRtextureNumRings(), 207
pfIBRtextureRing(), 207
pficonv, 241
pfiInit(), 733
pfImageCache, 386
pfInitBoard(), 594
pfIsBoardInit(), 594
pfIsPVChanActive(), 483
pfLayer, 70
pfLoadBalanceBalance(), 372
pfLoadBalanceLoad(), 372
pfLoadBalanceNumActive(), 372
pfLoadBalanceVal(), 372
pfLoadGState(), 340
pfLOD nodes, 70
pfLODRangeFlux(), 148

007-1680-080

Index

pfLODUserEvalFunc, 147
pfMatrix, 613
pfMatrix4d, 64
pfMatStack, 619
pfMPClipTexture, 415
connecting to pfPipes, 416
pfMPKImportConfig(), 46
pfMPKImportFile(), 45, 46, 47
pfMPKPostConfig(), 47
pfMPKPostDMConfig(), 47
pfMPKPreConfig(), 47
pfMQueryWin(), 478
pfMultithread(), 160, 174
pfMultithreadParami(), 174, 175, 176
pfNewLModel(), 330
pfNewPVChan(), 35, 483
PFNFYLEVEL environment variable, 503
pfNode, 52
and bounding volumes, 620
pfNode data structures, 49
attributes, 52
operations, 52
pfObject data structures, 6
actual type of, 17
pfOpenPWin(), 478
pfPartition, 78
pfPath data structures, 100
PFPATH environment variable, 504, 505
pfPipe
configuration, 23
data structures. See pipelines
pfPipeScreen(), 35, 483
pfPipeSwapFunc(), 476
pfPipeVideoChannel, 482
pfPlane, 622
pfPrint, 15

007-1680-080

pfProcessHighestPriority, 165
pfProcessPriorityUpgrade, 165
pfPVChanDVRMode(), 148
pfPVChanId(), 35, 483
pfPVChanMaxDecScale, 150
pfPVChanMaxIncScale, 151
pfPVChanMinDecScale, 151
pfPVChanMinIncScale, 151
pfPVChanMode(), 150
pfPVChanOutputAreaScale(), 485
pfPVChanOutputOrigin(), 485
pfPVChanOutputSize(), 485
pfPVChanStress, 151
pfPVChanStress(), 151
pfPVChanStressFilter, 151
pfPVChanStressFilter(), 151
pfPWinAddPVChan, 484
pfPWinAttach(), 481
pfPWinRemovePVChan, 484
pfPWinRemovePVChanIndex(), 484
pfPWinShare(), 481
pfPWinType(), 481
pfQueryWin(), 478
pfQueue, 487
pfQuickCopyGSet(), 178
pfQuickResetGSet(), 179
pfRemoveMPClipTexture(), 417
pfScene nodes, 27
pfSCS, 62, 64
pfSCS nodes, 658
pfSeg, 626
and bounding volumes, 620
pfSegSet
data structure, definition, 120
intersection with, 628
pfSelectWin(), 461, 464

777

Index

pfSequence, 67
pfsFace, 549
pfShaderChanged(), 348
pfShaderTexApplyFunc(), 349, 350
pfShaderTexComputeFunc(), 349, 350
pfShaderTexCoordFunc(), 349, 350
pfShadow class, 199
pfSphere, 621
pfState data structures, 338
pfString, 307
pfSwitch, 67
pfSwitchValFlux(), 148
pfSync(), 638, 711
pfTEnvMode(), 328
pfTerrainAttr(), 555
pfTexAnisotropy(), 326
pfTexEnv data structures. See texturing
pfTexFormat(), 327
pfTexGen, 301
pfTexLOD, 328
pfTexture, 396
pfTexture data structures. See texturing
pfTGenPoint(), 329
pfuAddMPClipTexturesToPipes, 387
pfuAddMPClipTextureToPipes, 387, 393
pfuCalcSizeFinestMipLOD, 435
pfuCalcVirtualClipTexParams, 435
pfuChooseFBConfig(), 459
pfuClipCenterNode, 418
pfuClipTexConfig structure, 398
pfuDownloadTexList(), 323
pfuFindClipTextures, 387
pfuFreeClipTexConfig, 387
pfuFreeImgCacheConfig, 387
pfuImgCacheConfig, 398

778

pfuInit(), 733
pfuInitClipTexConfig, 387
pfuInitImgCacheConfig, 387
pfuMakeClipTexture, 387
pfuMakeImageCache, 387
pfuMakeSceneTexList(), 323
pfUnbindPWinPVChans, 485
pfuNewClipCenterNode, 419
pfuProcessClipCenters, 387, 393, 418
pfuProcessClipCentersWithChannel, 387, 393, 418
pfvDisplayMngr class, 44
pfVec2, 611
pfVec3, 611
pfVec4, 611
pfVideoChannel, 34
pfVideoChannelInfo(), 35
pfview program, 703
pfvInteractor class, 712
pfvmDrawStyle module, 716
pfvmEarthSky module, 716
pfvmLoader module, 716
pfvmLogo module, 716
pfvmNavigator module, 701, 716
pfvmPicker module, 701, 716
pfvmSnapshot module, 717
pfvmStats module, 717
pfvmTrackball module, 717
pfvmWorldSwitcher module, 717
pfVolFog class, 186
pfvPicker class, 712
pfvSelector class, 712
pfvViewer class, 44, 700
pfWaitForVmeBus(), 602
pfWinShare(), 463
pfWinSwapBarrier(), 464

007-1680-080

Index

phase
defined, 751
PHD format. See formats
PHIGS, 257
picking, 100, 125
pipe, 351
pipe windows, 469
pipe, definition, 751
pipelines
functional stages, 21
multiple, 159, 495
multiprocessing, 115
overview, 21
pitch, 30
defined, 751
pixie, 673
point-volume intersection tests, 624
POLY format. See formats
polytopes, 103, 104
popping
definition, 752
in LOD transitions, 145
positive rotation, 30
postConfig(), 711
postCull(), 712
postDraw(), 712
preConfig(), 711
preCull(), 712
preDraw(), 712
previous statistics, 653
See also statistics
primitives
attributes, 299
connectivity, 296
flat-shaded, 294
types, 293
printing objects, 15

007-1680-080

process callbacks, 115
defined, 752
process priority, 165, 664
processor isolation, 664
prof, 673
profiling
prof, 673
program counter sampling, 673
programming modules, 700
projective texture
defined, 752
proto tile, 394
prune, definition, 752
pruning nodes, 90
PTU format. See formats
public structs, 720
punch through, definition, 752

Q
quaternion, 617
spherical linear interpolation, 617
use in C++ API, 720
query array, 559
queue, 487
retrieving elements, 489

R
r_streams, 407
rapid rendering, for on-air broadcast, xxxix
raster displays, 588
REACT, 664, 673
read function, 388
custom, 435, 448
sorting, 428

779

Index

read queue, 388, 428
read(), 502
ReadDirect, 436
ReadNormal, 436
RealityEngine graphics
pipelines, multiple, 159
tuning, 680
real-time programming, 664
real-time shadows, 198
reference count, definition, 752
reference counting, 11
reference point array, 553
reference position, 545
reference vertices, 550
refresh rate, 132
rendering
modes, 313
multiple channels, 35
stages of, 155
rendering pipelines
definition, 752
See pipelines
rendering values, 318
reserved functions, 720
right-hand rule, 30
defined, 753
roll, 30
defined, 753
rotating geometry to track eyepoint, 75, 671
rotations
quaternion, 617
rountines
pfShaderTexComputeFunc(), 350
routines, 611
enterView(), 712
enterWorld(), 712
exitView(), 712

780

exitWorld(), 712
for 3-Vectors, 612
for 4x4 Matrices, 613
for quatermions, 618
frame(), 711
handleEvent(), 709, 711
libpr
pfCompositorAddChild(), 368
pfCompositorChildFrustum(), 370
pfCompositorChildSubdivision(), 370
pfCompositorChildViewport(), 370
pfCompositorLoadBalancer(), 369
pfCompositorMasterPipe(), 370
pfCompositorMode(), 369
pfCompositorNearFar(), 368
pfCompositorOrtho(), 368
pfCompositorPerspective(), 368
pfCompositorReconfig(), 368
pfCompositorVal(), 369
pfCompositorViewport(), 368
pfGetCompositorChildFrustum(), 370
pfGetCompositorChildSubdivision(), 370
pfGetCompositorChildViewport(), 370
pfGetCompositorFrustType(), 368
pfGetCompositorFrustum(), 368
pfGetCompositorLoadBalancer(), 369
pfGetCompositorMode(), 369
pfGetCompositorNearFar(), 368
pfGetCompositorVal(), 369
pfGetCompositorViewport(), 368
pfGetLoadBalanceLoad(), 372
pfGetLoadBalanceNumActive(), 372
pfGetLoadBalanceVal(), 372
pfGetLoadBalanceWork(), 372
pfLoadBalanceBalance(), 372
pfLoadBalanceLoad(), 372
pfLoadBalanceNumActive(), 372
pfLoadBalanceVal(), 372
matrix stack, 619
overlay(), 712
pfAccumulateStats(), 648

007-1680-080

Index

pfAddChan(), 44
pfAddChild(), 54, 164, 231
pfAddGSet(), 12, 72, 73
pfAddMat(), 614
pfAddScaledVec3(), 612
pfAddVec3(), 612
pfAllocChanData(), 117, 170, 664
pfAllocIsectData(), 170
pfAlmostEqualMat(), 615
pfAlmostEqualVec3(), 613
pfAlphaFunc(), 311, 315, 671, 672
pfAlphaFunction(), 681
pfAntialias(), 318, 473, 657, 662, 719
pfApp(), 168
pfAppearanceChanged(), 348
pfAppFrame(), 88
pfApplyCtab(), 319, 332
pfApplyFog(), 319
pfApplyGState(), 312, 340, 341, 342, 343
pfApplyGStateTable(), 343
pfApplyHlight(), 319, 333
pfApplyLModel(), 319
pfApplyLPState(), 319
pfApplyMtl(), 167, 319
pfApplyTEnv(), 319, 320
pfApplyTex(), 167, 312, 319, 323, 324, 340, 666
pfApplyTGen(), 319, 329
pfApplyVolFog(), 187, 188, 193
pfAsynchDelete(), 164
pfAttachChan(), 40
pfAttachDPool(), 499
pfAttachPWin(), 362, 472
pfAttachPWinSwapGroup(), 362
pfAttachPWinWin(), 361
pfAttachPWinWinSwapGroup(), 362
pfAttachWin(), 463
pfAverageStats(), 648
pfBboardAxis(), 75
pfBboardMode(), 76
pfBboardPos(), 75
pfBeginSprite(), 335, 336

007-1680-080

pfBindPWinPVChans(), 361
pfBoxAroundBoxes(), 622
pfBoxAroundPts(), 622
pfBoxAroundSpheres(), 622
pfBoxContainsBox(), 625
pfBoxContainsPt(), 624
pfBoxExtendByBox(), 623
pfBoxExtendByPt(), 623
pfBoxIsectSeg(), 627
pfBufferAddChild(), 164, 739
pfBufferClone(), 164, 739
pfBufferRemoveChild(), 164, 739
pfBuildPart(), 78, 79
pfCBufferChanged(), 501
pfCBufferConfig(), 499, 502
pfCBufferFrame(), 501, 502
pfChanBinOrder(), 99, 738
pfChanBinSort(), 99, 738
pfChanESky(), 27, 181, 185
pfChanFOV(), 29
pfChanGState(), 658
pfChanLODAttr(), 131
pfChanLODLODStateIndex(), 141
pfChanLODStateList(), 141
pfChanNearFar(), 30
pfChanNodeIsectSegs(), 120
pfChanPick(), 125
pfChanScene(), 27, 61
pfChanShare(), 41, 88, 472
pfChanStatsMode(), 645
pfChanStress(), 131
pfChanStressFilter(), 131, 154
pfChanTravFunc(), 100, 117, 171
pfChanTravFuncs(), 181
pfChanTravMask(), 112
pfChanTravMode(), 93, 102, 103, 110, 659, 669
pfChanView(), 30, 32
pfChanViewMat(), 31, 32
pfChanViewOffsets(), 40
pfChooseFBConfig(), 459
pfChoosePWinFBConfig(), 362, 473, 478

781

Index

pfChooseWinFBConfig(), 459
pfClear(), 167
pfClearChan(), 118, 181, 639, 662
pfClearStats(), 648
pfClipSeg(), 626, 628
pfClipTextureAllocatedLevels(), 394
pfClipTextureClipSize(), 394
pfClipTextureEffectiveLevels(), 394
pfClipTextureInvalidBorder(), 394
pfClipTextureLevel(), 394, 397
pfClipTextureVirtualSize(), 394
pfClockMode(), 494
pfClockName(), 494
pfClone(), 164
pfCloseDList(), 337
pfCloseFile(), 502
pfClosePWin(), 478, 480
pfClosePWinGL(), 478
pfCloseWin(), 461
pfCloseWinGL(), 463
pfCombineVec3(), 612
pfCompositorChannelClipped(), 368
pfConfig(), 23, 46, 47, 160, 163, 502, 509, 675, 711,
733, 734
pfConfigPWin(), 476, 480, 666
pfConfigStage(), 24, 666
pfConjQuat(), 618
pfCopy(), 15, 305, 497, 651
pfCopyFStats(), 647, 652
pfCopyGSet(), 292
pfCopyGState(), 342
pfCopyMat(), 614
pfCopyStats(), 647, 648, 652
pfCopyVec3(), 612
pfCreateDPool(), 498, 499
pfCreateFile(), 502
pfCrossVec3(), 612
pfCull(), 117, 158, 168, 639, 712
pfCullFace(), 317, 657
pfCullPath(), 100
pfCullProgramAddBinParent(), 109

782

pfCullProgramAddPgmInstruction(), 105
pfCullProgramAddPgmOpcode(), 105
pfCullProgramIsSubbinOf(), 109
pfCullProgramResetBinParents(), 109
pfCullProgramResetPgm(), 105
pfCullProgramTestPolytope(), 109
pfCullResult(), 113
pfCurCBufferIndex(), 499
pfCylAroundSegs(), 622, 667
pfCylContainsPt(), 624
pfdAddExtAlias(), 226
pfDBase(), 165, 168
pfDBaseFunc(), 163, 660
pfdBldrStateAttr(), 233
pfdBldrStateMode(), 233
pfdBldrStateVal(), 233
pfdCleanTree(), 232, 658, 668, 670
pfdConverterAttr(), 226
pfdConverterAttr_rpc(), 273
pfdConverterMode(), 226
pfdConverterMode_rpc(), 273
pfdConverterVal(), 226
pfdConverterVal_rpc(), 273
pfdConvertFrom(), 223
pfdConvertTo(), 223
pfDCSCoord(), 63
pfDCSMat(), 63
pfDCSRot(), 63, 527
pfDCSScale(), 63, 526, 527
pfDCSScaleXYZ(), 526, 527
pfDCSTrans(), 63, 526, 527
pfdDefaultGState(), 658
pfDecal(), 316, 340, 657, 663, 744
pfDelete(), 11, 13, 165, 292, 305, 342, 497, 499
datapools, 499
pfDetachChan(), 40
pfDetachPWinSwapGroup(), 362
pfDetachPWinWin(), 362
pfdExitConverter(), 226
pfdFreezeTransforms(), 658, 668, 670
pfdGetConverterAttr(), 226

007-1680-080

Index

pfdGetConverterAttr_rpc(), 273
pfdGetConverterMode(), 226
pfdGetConverterMode_rpc(), 273
pfdGetConverterVal(), 226
pfdGetConverterVal_rpc(), 273
pfdInitConverter(), 226, 733
pfDisable(), 167, 318
pfDistancePt3(), 612
pfDivQuat(), 618
pfdLoadBldrState(), 233
pfdLoadClipTexture(), 400
pfdLoadClipTextureState(), 400
pfdLoadFile(), 223, 225, 228, 238, 504
pfdLoadImageCache(), 398
pfdLoadImageTileFormat(), 398
pfdMakeSceneGState(), 61, 658
pfdMakeSharedScene(), 61, 658, 668
pfdOptimizeGStateList(), 61, 658
pfDotVec3(), 612
pfDPoolAlloc(), 499
pfDPoolAttachAddr(), 499
pfDPoolFind(), 499
pfDPoolLock(), 499
pfDPoolUnlock(), 499
pfdPopBldrState(), 233
pfdPushBldrState(), 233
pfDraw(), 117, 158, 168, 639, 662, 665, 671, 712
pfDrawChanStats(), 635, 645, 647, 659, 669, 672,
676
pfDrawDList(), 304, 337, 340
pfDrawFStats(), 635, 645, 647
pfDrawGSet(), 167, 292, 294, 295, 304, 340, 342
pfDrawString(), 167, 307, 309
pfDrawVolFog(), 188, 191
pfdSaveBldrState(), 233
pfdStoreFIle(), 223
pfEarthSky(), 118
pfEnable(), 167, 318, 340
pfEnableStatsHw(), 645, 646, 649
pfEndSprite(), 336
pfEngineDst(), 530

007-1680-080

pfEngineEvaluate(), 514, 531
pfEngineEvaluationRange(), 531
pfEngineIterations(), 530
pfEngineMask(), 530
pfEngineMode(), 523, 526, 528, 531
pfEngineSrc(), 530
pfEngineSrcChanged(), 514
pfEngineUserFunction(), 529
pfEqualMat(), 615
pfEqualVec3(), 613
pfESkyAttr(), 184
pfESkyColor(), 184
pfESkyFog(), 184
pfESkyMode(), 184, 681
pfEvaluateLOD(), 147
pfExpQuat(), 618
pfFCSFlux(), 532
pfFeature(), 663, 672
pfFilePath(), 504
pfFilePathv(), 504
pfFindFile(), 504
pfFlatten(), 62, 164, 232, 658, 661, 668, 670
pfFlattenString(), 309
pfFlushState(), 342
pfFluxCallDataFunc(), 509, 510
pfFluxDefaultNumBuffers(), 509
pfFluxDisableSyncGroup(), 516
pfFluxedGSetInit(), 518
pfFluxEnableSyncGroup(), 516
pfFluxEvaluate(), 514, 531
pfFluxEvaluateEye(), 514
pfFluxFrame(), 510
pfFluxInitData(), 509
pfFluxMask(), 514
pfFluxMode(), 512
pfFluxSyncComplete(), 516
pfFluxSyncGroup(), 515
pfFluxSyncGroupReady(), 516, 517
pfFluxWriteComplete(), 513, 517
pfFogRange(), 333
pfFogType(), 333

783

Index

pfFontAttr(), 306
pfFontCharGSet(), 305
pfFontCharSpacing(), 306
pfFontMode(), 306
pfFrame(), 88, 90, 115, 117, 118, 131, 158, 159, 163,
171, 470, 476, 502, 638, 659, 711, 719
pfFrameRate(), 130, 131, 636
pfFree(), 496, 497
pfFrustContainsBox(), 625
pfFrustContainsCyl(), 625
pfFrustContainsPt(), 624
pfFrustContainsSphere(), 625
pfFSatsClass(), 647
pfFStatsAttr(), 653
pfFStatsClass(), 644, 650
pfFStatsCountNode(), 647, 649
pfFullXformPt3(), 612
pfGeoSetIsectMask(), 513
pfGetArena(), 497
pfGetBboardAxis(), 75
pfGetBboardMode(), 76
pfGetBboardPos(), 75
pfGetChanFStats(), 635, 647
pfGetChanLoad(), 131
pfGetChanView(), 32
pfGetChanViewMat(), 32
pfGetChanViewOffsets, 32
pfGetCompositorChannelClipped(), 368
pfGetCullResult(), 113
pfGetCurGState(), 343
pfGetCurWSConnection(), 461, 464
pfGetDCSMat(), 63
pfGetEngineDst(), 530
pfGetEngineFunction(), 523
pfGetEngineIterations(), 530
pfGetEngineMask(), 530
pfGetEngineMode(), 523
pfGetEngineNumSrcs(), 530
pfGetEngineSrc(), 530
pfGetEngineUserFunction(), 529
pfGetFFlux(), 511

784

pfGetFilePath(), 504
pfGetFileStatus(), 502
pfGetFluxClientEngine(), 513
pfGetFluxCurData(), 511, 512, 520
pfGetFluxDataSize(), 509
pfGetFluxDefaultNumBuffers(), 509
pfGetFluxEnableSyncGroup(), 516
pfGetFluxFrame(), 528, 529
pfGetFluxMask(), 514
pfGetFluxMemory(), 510
pfGetFluxNamedSyncGroup(), 515
pfGetFluxNumClientEngines(), 514
pfGetFluxNumNamedSyncGroups(), 516
pfGetFluxNumSrcEngines(), 514
pfGetFluxSrcEngine(), 513
pfGetFluxSyncGroup(), 515
pfGetFluxSyncGroupName(), 515
pfGetFluxWritableData(), 512
pfGetGSet(), 72, 73
pfGetGSetPrimLength(), 292
pfGetIBRnodeProxyTexCoords(), 206
pfGetIBRtextureDefaultTexture(), 205
pfGetImageTileMemInfo (), 395
pfGetLayerBase(), 71
pfGetLayerDecal(), 71
pfGetLayerMode(), 71
pfGetLODCenter(), 70
pfGetLODRange(), 70
pfGetMatCol(), 614
pfGetMatColVec3(), 614
pfGetMatRow(), 614
pfGetMatRowVec3(), 614
pfGetMStack(), 620
pfGetMStackDepth(), 620
pfGetMStackTop(), 620
pfGetNumChildren(), 54
pfGetNumGSets(), 72, 73
pfGetOrthoMatCoord(), 614
pfGetOrthoMatQuat(), 614
pfGetParent(), 100
pfGetParentCullResult(), 113

007-1680-080

Index

pfGetPartAttr(), 79
pfGetPartType(), 79
pfGetPipe(), 23, 355
pfGetPipeScreen(), 24
pfGetPipeSize(), 24
pfGetPWinCurOriginSize(), 478
pfGetQuatRot(), 618
pfGetRef(), 12
pfGetSCSMat(), 62
pfGetSemaArena(), 170, 496, 498
pfgetSemaArena(), 338
pfGetSeqDuration(), 68
pfGetSeqFrame(), 68
pfGetSeqInterval(), 68
pfGetSeqMode(), 68
pfGetSeqTime(), 67
pfGetSharedArena(), 169, 299, 496
pfGetSize(), 497
pfGetSwitchVal(), 67
pfGetTime(), 494
pfGetType(), 17
pfGetTypeName(), 17
pfGetVClock(), 495
pfGetVolFogTexture(), 188
pfGetWinCurOriginSize(), 456, 457
pfGetWinCurScreenOriginSize(), 456
pfGetWinFBConfig(), 461
pfGetWinGLCxt(), 461
pfGetWinOrigin(), 456
pfGetWinSize(), 456
pfGetWinWSDrawable(), 460
pfGetWinWSWindow(), 460
pfGSetAttr(), 12, 292, 299, 497, 532
pfGSetBBox(), 292
pfGSetBound(), 532
pfGSetDrawMode(), 292, 294, 295, 304
pfGSetGState(), 12, 292, 342
pfGSetGStateIndex(), 292
pfGSetHlight(), 12, 292, 333
pfGSetIsectMask(), 126, 292, 629
pfGSetIsectSegs(), 292, 305, 628, 629

007-1680-080

pfGSetLineWidth(), 292
pfGSetMultiAttr(), 292, 299
pfGSetNumPrims(), 292, 293, 294
pfGSetOptimize(), 179
pfGSetPntSize(), 292
pfGSetPrimLengths(), 292, 293, 294
pfGSetPrimType(), 292
pfGSetQuickAttr(), 178
pfGSetQuickMultiAttr(), 179
pfGSetQuickPrimLengths(), 179
pfGStateAttr(), 12, 341, 343
pfGStateFuncs(), 343
pfGStateInherit(), 340, 343
pfGStateMode(), 314, 341, 342
pfGStateVal(), 318, 343
pfHalfSpaceContainsBox(), 625
pfHalfSpaceContainsCyl(), 625
pfHalfSpaceContainsPt(), 624
pfHalfSpaceContainsSphere(), 625
pfHalfSpaceIsectSeg(), 627
pfHlightColor(), 334
pfHlightLineWidth(), 334
pfHlightMode(), 334
pfHlightNormalLength(), 334
pfHlightPntSize(), 334
pfHyperpipe(), 166, 352, 353
pfIBRnodeAngles(), 205
pfIBRnodeFlags(), 206
pfIBRnodeIBRtexture(), 205
pfIBRnodeProxyTexCoords(), 206
pfIBRtextureDirection(), 208
pfIBRtextureFlags(), 207, 208
pfIBRtextureIBRdirections(), 206
pfIBRtextureIBRtextures(), 207
pfIBRtextureLoadIBRtexture(), 207
pfIBRtextureNumRings(), 207
pfIBRtextureRing(), 207
pfIdleTex(), 323
pfImageCacheFileStreamServer(), 395
pfImageCacheImageSize(), 395
pfImageCacheMemRegionOrigin(), 395

785

Index

pfImageCacheMemRegionSize(), 395
pfImageCacheName(), 395
pfImageCacheProtoTile(), 396
pfImageCacheTex(), 395
pfImageCacheTexRegionOrigin(), 395
pfImageCacheTexRegionSize(), 395
pfImageCacheTexSize(), 395
pfImageCacheTileFileNameFormat(), 395
pfImageTileDefaultTile(), 396
pfImageTileFileImageFormat(), 397
pfImageTileFileImageType(), 395
pfImageTileFileName(), 395, 396
pfImageTileHeaderOffset(), 395, 397
pfImageTileMemImageFormat(), 396
pfImageTileMemImageType(), 395, 397
pfImageTileMemInfo(), 395
pfImageTileNumFileTiles(), 395
pfImageTileReadFunc(), 395
pfImageTileReadQueue(), 395, 396
pfImageTileSize(), 395, 397
pfIndex(), 639
pfInit(), 160, 169, 497, 733, 737
pfInitArenas(), 496, 497, 498, 733, 737
pfInitCBuffer(), 502
pfInitClock(), 494
pfInitGfx(), 460, 473
pfInitState(), 338
pfInitVClock(), 495
pfInsertChan(), 38, 44
pfInsertChild(), 54
pfInsertGSet(), 12, 72, 73
pfInvertAffMat(), 615
pfInvertFullMat(), 615
pfInvertIdentMat(), 615
pfInvertOrthoMat(), 615
pfInvertOrthoNMat(), 615
pfInvertQuat(), 618
pfIsectFunc(), 158, 171, 660
pfIsOfType(), 17
pfIsTexLoaded(), 323
pfLayer(), 744

786

pfLayerBase(), 71
pfLayerDecal(), 71
pfLayerMode(), 71
pfLengthQuat(), 618
pfLengthVec3(), 612
pfLightAtten(), 331
pfLightOn(), 167, 319, 331
pfLoadImageTile(), 397
pfLoadMatrix(), 335
pfLoadMStack(), 620
pfLoadState(), 338
pfLoadTex(), 323
pfLoadTexFile(), 320
pfLODCenter(), 70
pfLODLODState(), 140
pfLODLODStateIndex(), 140
pfLODRange(), 70
pfLODTransition(), 145
pfLODUserEvalFunc(), 147
pfLogQuat(), 618
pfMakeCoordMat(), 614
pfMakeEulerMat(), 613
pfMakeOrthoFrust(), 622
pfMakePerspFrust(), 622
pfMakePolarSeg(), 626
pfMakePtsSeg(), 626
pfMakeQuatMat(), 613
pfMakeRotMat(), 613
pfMakeRotOntoMat(), 613
pfMakeRotQuat(), 618
pfMakeScaleMat(), 613
pfMakeTransMat(), 613
pfMalloc(), 12, 169, 299, 496, 497, 498, 668
pfMergeBuffer(), 163, 164, 739
pfModelMat(), 336
pfMoveChan(), 38, 44
pfMovePWin(), 480
pfMPKImportConfig(), 46
pfMPKImportFile(), 45, 46, 47
pfMPKPostConfig(), 47
pfMPKPostDMConfig(), 47

007-1680-080

Index

pfMPKPreConfig(), 47
pfMQueryFStats(), 648, 652
pfMQueryHit(), 121, 122, 628
pfMQueryStats(), 648, 652
pfMtlColorMode(), 332, 657, 661, 682
pfMultipipe(), 159, 160, 353, 354
pfMultiprocess(), 23, 156, 157, 158, 160, 163, 509,
637, 660, 675, 719, 733
pfMultithread(), 159, 160, 174
pfMultithreadParami(), 174, 175, 176
pfMultMat(), 614
pfMultMatrix(), 167, 335
pfMultQuat(), 618
pfNegateVec3(), 612
pfNewBboard(), 75
pfNewBuffer(), 163, 739
pfNewCBuffer(), 502
pfNewChan(), 27
pfNewClipTexture(), 394
pfNewCtab(), 332
pfNewDCS(), 63
pfNewDList(), 337
pfNewDPool(), 498
pfNewESky(), 184
pfNewFlux(), 508
pfNewFog(), 332
pfNewFont(), 305
pfNewFrust(), 622
pfNewGeode(), 72, 73
pfNewGSet(), 292
pfNewGState(), 342
pfNewHlight(), 333
pfNewImageTile(), 395, 396
pfNewLayer(), 71
pfNewLight(), 330
pfNewLModel(), 330
pfNewLOD(), 70
pfNewMaterial(), 331
pfNewMStack(), 619
pfNewMtl(), 331
pfNewPart(), 79

007-1680-080

pfNewPath(), 100
pfNewPWin(), 470, 477
pfNewScene(), 61
pfNewSCS(), 62
pfNewSeq(), 67
pfNewState(), 338, 460
pfNewString(), 308
pfNewSwitch(), 67
pfNewTex(), 320
pfNewVolFog(), 188
pfNewWin(), 454
pfNodeBSphere(), 60
pfNodeIsectSegs(), 120, 122, 125, 158, 628, 667
pfNodeTravData(), 730
pfNodeTravFuncs(), 112, 730, 732, 736
pfNodeTravMask(), 112, 122, 126, 667
pfNormalizeVec3(), 612
pfNotify(), 496, 503
pfNotifyHandler(), 496, 503
pfNotifyLevel(), 503, 664, 676
pfOpenDList(), 337
pfOpenFile(), 502
pfOpenPWin(), 470, 476, 477, 478, 479, 480
pfOpenScreen(), 456, 464
pfOpenStats(), 649
pfOpenWin(), 454, 457, 460, 461, 462, 463
pfOpenWSConnection(), 465
pfOrthoXformCyl(), 623
pfOrthoXformFrust(), 623
pfOrthoXformPlane(), 623
pfOrthoXformSphere(), 623
pfOverride(), 313, 319, 339, 662
pfPartAttr(), 79
pfPassChanData(), 118, 171, 659, 664
pfPassIsectData(), 171
pfPhase(), 131, 133
pfPipeScreen(), 24, 356
pfPipeWSConnectionName(), 356
pfPlaneIsectSeg(), 628
pfPopMatrix(), 113, 335
pfPopMStack(), 620

787

Index

pfPopState(), 338
pfPositionSprite(), 336
pfPostMultMat(), 614
pfPostMultMStack(), 620
pfPostRotMat(), 615
pfPostRotMStack(), 620
pfPostScaleMat(), 615
pfPostScaleMStack(), 620
pfPostTransMat(), 615
pfPostTransMStack(), 620
pfPreMultMat(), 614
pfPreMultMStack(), 620
pfPreRotMat(), 615
pfPreRotMStack(), 620
pfPreScaleMat(), 615
pfPreTransMat(), 614
pfPrint(), 292, 305, 651
pfProcessHighestPriority, 165
pfProcessPriorityUpgrade(), 165
pfPushIdentMatrix(), 335
pfPushMatrix(), 113, 335
pfPushMStack(), 619
pfPushState(), 167, 338
pfPWinAddPVChan(), 361
pfPWinConfigFunc(), 476, 477, 478
pfPWinFBConfig(), 361, 473, 478
pfPWinFBConfigAttrs(), 361, 473, 478
pfPWinFBConfigData(), 361
pfPWinFBConfigId(), 361
pfPWinFullScreen(), 470, 471, 478
pfPWinGLCxt(), 361, 478
pfPWinIndex(), 474, 477
pfPWinList(), 361
pfPWinMode(), 474, 477
pfPWinOriginSize(), 470, 477, 478
pfPWinOverlayWin(), 361
pfPWinPVChan(), 361
pfPWinRemovePVChan(), 361
pfPWinRemovePVChanIndex(), 361
pfPWinScreen(), 361, 471, 478
pfPWinShare(), 478

788

pfPWinStatsWin(), 361
pfPWinSwapBarrier(), 361
pfPWinType(), 478
pfPWinWSConnectionName(), 361
pfPWinWSDrawable(), 361, 479
pfPWinWSWindow(, 361
pfPWinWSWindow(), 478, 479
pfQuatMeanTangent(), 618
pfQueryFeature(), 656
pfQueryFStats(), 648, 652
pfQueryGSet(), 292
pfQueryHit(), 121, 122, 628, 630
pfQueryStats(), 648, 652
pfQuerySys(), 459
pfQueryWin(), 459, 460
pfQuickCopyGSet(), 178
pfQuickResetGSet(), 179
pfReadFile(), 502
pfRef(), 12
pfReleaseDPool(), 499
pfRemoveChan(), 44
pfRemoveChild(), 54, 164
pfRemoveGSet(), 72, 73
pfReplaceGSet(), 12, 72, 73
pfResetDList(), 337
pfResetMStack(), 619
pfResetStats(), 648
pfRotate(), 335
pfScale(), 335
pfScaleVec3(), 612
pfSceneGState(), 61, 658
pfSeekFile(), 502
pfSegIsectPlane(), 628
pfSegIsectTri(), 628
pfSelectBuffer(), 163, 739
pfSelectPWin(), 361
pfSelectState(), 338
pfSelectWin(), 456
pfSelectWSConnection(), 464
pfSeqDuration(), 68
pfSeqInterval(), 68

007-1680-080

Index

pfSeqMode(), 68
pfSeqTime(), 67
pfSetMatCol(), 614
pfSetMatColVec3(), 614
pfSetMatRow(), 614
pfSetMatRowVec3(), 614
pfSetVec3(), 612
pfShadeModel(), 315
pfShaderChanged(), 348
pfShaderTexApplyFunc(), 349, 350
pfShaderTexComputeFunc(), 349
pfShaderTexCoordFunc(), 349, 350
pfShadow routines, 200
pfSharedArenaSize(), 496, 498
pfSlerpQuat(), 618
pfSphereAroundBoxes(), 622
pfSphereAroundPts(), 622
pfSphereAroundSpheres(), 622
pfSphereContainsCyl(), 625
pfSphereContainsPt(), 624
pfSphereContainsSphere(), 625
pfSphereExtendByPt(), 623
pfSphereExtendBySphere(), 623
pfSphereIsectSeg(), 627
pfSpriteAxis(), 336
pfSpriteMode(), 336
pfSqrDistancePt3(), 612
pfSquadQuat(), 618
pfStageConfigFunc(), 24
pfStatsClass(), 644, 649
pfStatsClassMode(), 648, 649, 650
pfStatsCountGSet(), 649
pfStatsHwAttr(), 644, 649
pfStringColor(), 309
pfStringFont(), 307
pfStringMat(), 309
pfStringMode(), 309
pfSubloadTex(), 323
pfSubloadTexLevel(), 323
pfSubMat(), 614
pfSubVec3(), 612

007-1680-080

pfSwapWinBuffers(), 364
pfSwitchVal(), 67
pfSync(), 88, 131, 638, 659, 711
pfTevMode(), 657
pfTexAnisotropy(), 326
pfTexDetail(), 12, 326
pfTexFilter(), 325, 657
pfTexFormat(), 323, 324, 394, 396, 681
pfTexFrame(), 324
pfTexImage(), 12, 320, 322, 394, 396
pfTexLevel(), 326
pfTexList(), 324
pfTexLoadImage(), 324
pfTexLoadMode(), 321, 324
pfTexLoadOrigin(), 322
pfTexLoadSize(), 324
pfTexName(), 394
pfTexSpline(), 326
pfTGenMode(), 329
pfTGenPlane(), 329
pfTranslate(), 335
pfTransparency(), 167, 311, 314, 340, 657, 662
pfTransposeMat(), 614
pfTriIsectSeg(), 628
pfuAddMPClipTexturesToPipes, 393
pfuCollideSetup(), 659
pfuDownloadTexList(), 364, 666
pfuFreeClipTexConfig(), 397
pfuFreeImgCacheConfig(), 398
pfuInitClipTexConfig(), 397
pfuInitImgCacheConfig(), 398
pfuLockDownApp(), 664
pfuLockDownCull(), 664
pfuLockDownDraw(), 664
pfuLockDownProc(), 664
pfuMakeClipTexture(), 397
pfuMakeImageCache(), 398
pfUnbindPWinPVChans(), 361
pfUnref(), 12
pfUnrefDelete(), 14
pfUpdatePart(), 78, 79

789

Index

pfuPrioritizeProcs(), 664
pfUserData(), 730
pfVClockSync(), 495
pfViewMat(), 336
pfVolFogAddChannel(), 188, 191
pfVolFogAddColoredPoint(), 187, 188, 192
pfVolFogAddNode(), 187, 188, 194
pfVolFogAddPoint(), 187, 188, 192
pfVolFogSetAttr(), 188, 192, 193
pfVolFogSetColor(), 188, 192
pfVolFogSetDensity(), 188
pfVolFogSetFlags(), 188, 193
pfVolFogSetVal(), 188, 192, 193, 195
pfVolFogUpdateView(), 188, 191
pfWinFBConfig(), 460, 461
pfWinFBconfig(), 459
pfWinFBConfigAttrs(), 457
pfWinFullScreen(), 456, 471
pfWinGLCxt(), 459, 460, 461
pfWinIndex(), 462, 463
pfWinMode(), 461
pfWinOriginSize(), 456
pfWinOverlayWin(), 463
pfWinScreen(), 456
pfWinShare(), 463
pfWinStatsWin(), 463
pfWinType(), 455
pfWinWSDrawable(), 459, 460
pfWinWSWindow(), 460, 462
pfWriteFile(), 502
pfXformBox(), 623
pfXformPt3(), 612
pfXformVec3(), 612
postConfig(), 711
postCull(), 712
postDraw(), 712
preConfig(), 711
preCull(), 712
preDraw(), 712
sync(), 711
RPC database format, 273

790

S
s_streams, 407
sample code, 716
sample code. See example code
sampling, program counter, 673
Scalable Graphics Compositor, 351, 365
scan rate, 132
scene complexity, definition, 753
scene graph
defined, 753
state inheritance, 88
scene graphs, 87
scene, definition, 753
screen-door transparency, 315
SCS. See pfSCS nodes
search paths, 504
definition, 753
segments, 626
See also pfSegSet
self-shadowing, 195
semaphores, allocating, 498
sense, definition of, 753
setmon(), 495
setrlimit(), 498
setSyncGroup, 559
SGF database format, 275
SGO format. See formats
Shader
flow through Performer, 350
man pages, 347
shader, 347
shading
flat, 315
Gouraud, 315
shadow, 559

007-1680-080

Index

shadow map
defined, 753
shadows, 198
share groups, 422, 754
share mask, 388, 422, 754
shared arena, memory mapping, 497
shared instancing, 56
defined, 754
shared memory
allocation, 726
arenas, 496
datapools, 498
debugging and, 675
sharing channel attributes, 40
sharpen texture, 671
shininess, definition, 754
siblings, of a node, defined, 754
Sierpinski sponge, 280
Silicon Graphics Object format. See formats
Simplify application, 209
simulation, xxxix
simulation-based design, xxxix
single inheritance, 10
single-precision arithmetic, 672
slave cliptexture, 420
slew table, 596
smallest_icache, 411
Software Systems, 252
Soma cube puzzle, 249
sorting, 490
defined, 754
sorting for transparency, 315
source code
See also example code
database loaders, 230, 730
database loading, 225
database processing tools, 221

007-1680-080

fill statistics, 644
libpfutil, 664, 666
MIPmap, 321
Perfly, 751
pfPipeWindow, 474, 479, 480
pfTexture, 321
statistics, 645, 654
X input, 666
spacing
character, 305
definition, 754
spatial decomposition, 352
spatial organization, 94
definition, 755
SPF format. See formats
spheres
as bounding volumes, 621
SPONGE format. See formats
sprite, 335
defined, 755
sproc(), 168, 502, 667, 733
stack, 725
stage timing graph, 636, 637
See also statistics
stage, definition, 755
stages of rendering, 155
STAR format. See formats
state
changes, 657
defined, 755
inheritance, 87
local and global, 340
state elements, 311
state specification
global, 340
local, 340
static coordinate systems. See pfSCS nodes
static data in C++ classes, 734

791

Index

statistics, 635
average, 653
CPU, 641
cumulative, 653
current, 653
data structures, 635, 653
displaying, 635, 636, 645
enabling, 647
fill, 644
graphics, 643
previous, 653
stage timing
defaults, 646
graph, 637
use in applications, 645
stencil decaling, 316
defined, 755
stereo display, 38
STL format. See formats
stream, 407
stress filter, 151
stress filter for DVR, 151
stress management, 152
stress management. See load management
stress, definition, 755
structures
libpfdu
pfdBuilder, 256
sub-bin, 98
subclassing, 729
subgraph, definition, 755
supersampled data, 439
SuperViewer, 283
SV format. See formats
switch nodes, 67
sync(), 711
synchronization of frames, 132

792

T
t_streams, 407
tearing, 316
temporal decomposition, 352
testing
intersections. See intersections
visibility, 93
tex region, 376
tex_region_size, 404
texel coordinates, 374
texel data, 384, 391
texel format, 394
texel tile, 391
texel, definition, 756
texload time, 427
text, 73
texture
coordinate generation, 329
coordinates, 554
detail, 671
loading, 328
magnification, 671
mapping, defined, 756
memory, 379
minification, 671
paging, 323
sharpen, 671
tiling, 391
texturing
overview, 320
performance cost, 661, 680
RealityEngine graphics, 681
representing complex objects, 671
tile size, 392
tile, algorithm, 390
tile, default, 396
tile, defined, 756

007-1680-080

Index

tile, updates, 434
tile_base, 413
tile_files, 412
tile_format, 405, 413
tile_params, 405, 413
tile_size, 413
tiles, 375
tiles_in_file, 413
tiling an image, 391
tiling, strategy, 433
TIN, 538
tokens
APP_CULL_DRAW, 673
PF_MAX_ANISOTROPY, 326
PF_MAX_LIGHTS, 331
PF_OFF, 314
PFAA_OFF, 314
PFAF_ALWAYS, 313
PFAF_GREATER, 315
PFBOUND_FLUX, 532
PFBOUND_STATIC, 60
PFCF_BACK, 317
PFCF_BOTH, 317
PFCF_FRONT, 317
PFCF_OFF, 314, 317
PFCHAN_EARTHSKY, 41
PFCHAN_FOV, 41
PFCHAN_LOD, 41
PFCHAN_NEARFAR, 41
PFCHAN_SCENE, 41
PFCHAN_STRESS, 41
PFCHAN_SWAPBUFFERS, 41
PFCHAN_SWAPBUFFERS_HW, 464, 482
PFCHAN_VIEW, 41
PFCHAN_VIEW_OFFSETS, 41
PFCULL_GSET, 102
PFCULL_IGNORE_LSOURCES, 102
PFCULL_PROGRAM, 102
PFCULL_SORT, 102, 659

007-1680-080

PFCULL_VIEW, 102
PFDECAL_BASE_STENCIL, 316, 317
PFDECAL_LAYER_STENCIL, 317
PFDECAL_OFF, 314, 317
PFDL_RING, 337
PFDRAW_OFF, 103
PFDRAW_ON, 103
PFEN_COLORTABLE, 318
PFEN_FOG, 318
PFEN_HIGHLIGHTING, 319
PFEN_LIGHTING, 318
PFEN_LPOINTSTATE, 319
PFEN_TEXGEN, 319
PFEN_TEXTURE, 318
PFEN_WIREFRAME, 318
PFENG_ALIGN, 522
PFENG_ANIMATE, 522
PFENG_ANIMATE_BASE_MATRIX, 523
PFENG_ANIMATE_ROT, 523
PFENG_ANIMATE_SCALE_UNIFORM, 523
PFENG_ANIMATE_SCALE_XYZ, 523
PFENG_ANIMATE_TRANS, 523
PFENG_BBOX, 522
PFENG_BLEND, 522
PFENG_MATRIX, 522
PFENG_MORPH, 522
PFENG_MORPH_FRAME, 524
PFENG_MORPH_SRC(n), 524
PFENG_MORPH_WEIGHTS, 524
PFENG_RANGE_CHECK, 531
PFENG_STROBE, 523
PFENG_TIME, 522
PFENG_TRANSFORM, 522
PFENG_USER_FUNCTION, 523
PFES_BUFFER_CLEAR, 182
PFES_FAST, 184
PFES_GRND_FAR, 182
PFES_GRND_HT, 182
PFES_GRND_NEAR, 182
PFES_SKY, 185
PFES_SKY_CLEAR, 185

793

Index

PFES_SKY_GRND, 182, 185
PFFB_ACCUM_ALPHA_SIZE, 458
PFFB_ACCUM_BLUE_SIZE, 458
PFFB_ACCUM_GREEN_SIZE, 458
PFFB_ACCUM_RED_SIZE, 458
PFFB_ALPHA_SIZE, 458
PFFB_AUX_BUFFER, 458
PFFB_BLUE_SIZE, 458
PFFB_BUFFER_SIZE, 458
PFFB_DEPTH_SIZE, 458
PFFB_DOUBLEBUFFER, 458
PFFB_GREEN_SIZE, 458
PFFB_RED_SIZE, 458
PFFB_RGBA, 458
PFFB_STENCIL, 458
PFFB_STEREO, 458
PFFB_USE_GL, 458
PFFLUX_BASIC_MASK, 514
PFFLUX_DEFAULT_NUM_BUFFERS, 509
PFFLUX_ON_DEMAND, 512, 520
PFFOG_PIX_EXP, 333
PFFOG_PIX_EXP2, 333
PFFOG_PIX_LIN, 333
PFFOG_PIX_SPLINE, 333
PFFOG_VTX_EXP, 333
PFFOG_VTX_EXP2, 333
PFFOG_VTX_LIN, 333
PFFONT_BBOX, 306
PFFONT_CHAR_SPACING, 306
PFFONT_CHAR_SPACING_FIXED, 306
PFFONT_CHAR_SPACING_VARIABLE, 306
PFFONT_GSTATE, 306
PFFONT_NAME, 306
PFFONT_NUM_CHARS, 306
PFFONT_RETURN_CHAR, 306
PFFONT_SPACING, 306
PFGS_COLOR4, 332
PFGS_COMPILE_GL, 295, 304
PFGS_COORD3, 513
PFGS_FLAT_LINESTRIPS, 293, 302
PFGS_FLAT_TRISTRIPS, 293, 302

794

PFGS_FLATSHADE, 294
PFGS_LINES, 293
PFGS_LINESTRIPS, 293
PFGS_OFF, 301
PFGS_OVERALL, 301
PFGS_PER_PRIM, 301
PFGS_PER_VERTEX, 301, 302
PFGS_POINTS, 293
PFGS_POLYS, 293
PFGS_QUADS, 293, 629
PFGS_TRIS, 293, 629
PFGS_TRISTRIPS, 293, 629
PFGS_WIREFRAME, 294
PFHL_BBOX_FILL, 334
PFHL_BBOX_LINES, 334
PFHL_FILL, 334
PFHL_FILL_R, 334
PFHL_FILLPAT, 334
PFHL_FILLPAT2, 334
PFHL_FILLTEX, 334
PFHL_LINES, 334
PFHL_LINES_R, 334
PFHL_LINESPAT, 334
PFHL_LINESPAT2, 334
PFHL_NORMALS, 334
PFHL_POINTS, 334
PFHL_SKIP_BASE, 334
PFIS_ALL_IN, 113, 625, 627
PFIS_FALSE, 113, 624, 625, 628
PFIS_MAYBE, 625, 627
PFIS_PICK_MASK, 125
PFIS_START_IN, 627
PFIS_TRUE, 624, 625, 627, 628
PFLUX_WRITE_ONCE, 513
PFMP_APP_CULL_DRAW, 157, 159, 167, 169, 637
PFMP_APP_CULLDRAW, 156, 158, 167
PFMP_APPCULL_DRAW, 157, 159
PFMP_APPCULLDRAW, 156, 158, 159, 675
PFMP_CULL_DL_DRAW, 157, 158, 658, 659
PFMP_CULLoDRAW, 157, 637
PFMP_FORK_COMPUTE, 158

007-1680-080

Index

PFMP_FORK_CULL, 156, 157
PFMP_FORK_DBASE, 163
PFMP_FORK_DRAW, 156, 157
PFMP_FORK_ISECT, 158
PFMTL_CMODE_AD, 661
PFNFY_ALWAYS, 503
PFNFY_DEBUG, 225, 503, 676
PFNFY_FATAL, 496, 503
PFNFY_FP_DEBUG, 503
PFNFY_INFO, 503
PFNFY_NOTICE, 503
PFNFY_WARN, 503
PFPB_LEVEL, 458
PFPHASE_FLOAT, 133
PFPHASE_FREE_RUN, 133, 471
PFPHASE_LIMIT, 133
PFPHASE_LOCK, 133
PFPK_M_ALL, 125
PFPK_M_NEAREST, 125
PFPROC_APP, 24
PFPROC_CULL, 24
PFPROC_DBASE, 24
PFPROC_DRAW, 24
PFPROC_ISECT, 24
PFPVC_DVR_AUTO, 148, 150, 485
PFPVC_DVR_MANUAL, 150, 485
PFPWIN_TYPE_NONEVENTS, 471
PFPWIN_TYPE_NOPORT, 456
PFPWIN_TYPE_OVERLAY, 455
PFPWIN_TYPE_SHARE, 472, 481
PFPWIN_TYPE_STATS, 455, 472
PFPWIN_TYPE_X, 455, 472
PFQHIT_FLAGS, 121
PFQHIT_GSET, 121
PFQHIT_NAME, 121
PFQHIT_NODE, 121
PFQHIT_NORM, 121
PFQHIT_PATH, 121
PFQHIT_POINT, 121
PFQHIT_PRIM, 121
PFQHIT_SEG, 121

007-1680-080

PFQHIT_SEGNUM, 121
PFQHIT_TRI, 121
PFQHIT_VERTS, 121
PFQHIT_XFORM, 121
PFSM_FLAT, 315
PFSM_GOURAUD, 314, 315
PFSORT_BACK_TO_FRONT, 99
PFSORT_BY_STATE, 99
PFSORT_END, 99
PFSORT_FRONT_TO_BACK, 99
PFSORT_NO_ORDER, 99
PFSORT_QUICK, 99
PFSORT_STATE_BGN, 99
PFSORT_STATE_END, 99
PFSPRITE_AXIAL_ROT, 336
PFSPRITE_MATRIX_THRESHOLD, 336
PFSPRITE_POINT_ROT_EYE, 336
PFSPRITE_POINT_ROT_WORLD, 336
PFSPRITE_ROT, 336
PFSTATE_ALPHAFUNC, 313
PFSTATE_ALPHAREF, 318
PFSTATE_ANTIALIAS, 314
PFSTATE_BACKMTL, 319
PFSTATE_COLORTABLE, 319
PFSTATE_CULLFACE, 314
PFSTATE_DECAL, 314
PFSTATE_ENCOLORTABLE, 314
PFSTATE_ENFOG, 314
PFSTATE_ENHIGHLIGHTING, 314
PFSTATE_ENLIGHTING, 314, 341
PFSTATE_ENLPOINTSTATE, 314
PFSTATE_ENTEXGEN, 314
PFSTATE_ENTEXTURE, 314, 341
PFSTATE_ENWIREFRAME, 314
PFSTATE_FOG, 319
PFSTATE_FRONTMTL, 319
PFSTATE_HIGHLIGHT, 319
PFSTATE_LIGHTMODEL, 319
PFSTATE_LIGHTS, 319
PFSTATE_LPOINTSTATE, 319
PFSTATE_SHADEMODEL, 314

795

Index

PFSTATE_TEXENV, 319
PFSTATE_TEXGEN, 319
PFSTATE_TEXTURE, 319
PFSTATE_TRANSPARENCY, 311, 313
PFSTATS_ENGFX, 643
PFSTATS_ON, 643
PFSTR_CENTER, 309
PFSTR_CHAR, 309
PFSTR_CHAR_SIZE, 309
PFSTR_FIRST, 309
PFSTR_INT, 309
PFSTR_JUSTIFY, 309
PFSTR_LAST, 309
PFSTR_LEFT, 309
PFSTR_MIDDLE, 309
PFSTR_RIGHT, 309
PFSTR_SHORT, 309
PFSWITCH_OFF, 67
PFSWITCH_ON, 67
PFTEX_BASE_APPLY, 325
PFTEX_BASE_AUTO_REPLACE, 324
PFTEX_BASE_AUTO_SUBLOAD, 325
PFTEX_FAST, 325
PFTEX_LIST_APPLY, 325
PFTEX_LIST_AUTO_IDLE, 325
PFTEX_LIST_AUTO_SUBLOAD, 325
PFTEX_LOAD_BASE, 325
PFTEX_LOAD_LIST, 325
PFTEX_LOAD_SOURCE, 321
PFTEX_SOURCE_FRAMEBUFFER, 322
PFTEX_SOURCE_IMAGE, 322
PFTEX_SOURCE_VIDEO, 322
PFTEX_SUBLOAD_FORMAT, 323
PFTG_EYE_PLANE, 329
PFTG_EYE_PLANE_IDENT, 329
PFTG_OBJECT_PLANE, 329
PFTG_SPHERE_MAP, 330
PFTR_BLEND_ALPHA, 315
PFTR_FAST, 314
PFTR_HIGH_QUALITY, 315
PFTR_MS_ALPHA, 315

796

PFTR_NO_OCCLUDE, 315
PFTR_OFF, 313, 314
PFTR_ON, 314
PFTRAV_CONT, 112, 124, 630
PFTRAV_CULL, 102, 117
PFTRAV_DRAW, 117
PFTRAV_IS_BCYL, 126
PFTRAV_IS_CACHE, 659
PFTRAV_IS_CLIP_END, 125, 630
PFTRAV_IS_CLIP_START, 125, 630
PFTRAV_IS_CULL_BACK, 126
PFTRAV_IS_GEODE, 124
PFTRAV_IS_GSET, 124, 126, 629
PFTRAV_IS_IGNORE, 124, 629
PFTRAV_IS_NO_PART, 78
PFTRAV_IS_NODE, 126
PFTRAV_IS_PRIM, 124, 629
PFTRAV_IS_UNCACHE, 513
PFTRAV_PRUNE, 112, 113, 124, 630
PFTRAV_TERM, 112, 113, 124, 630
PFVCHAN_AUTO_APPLY, 485
PFVCHAN_SYNC, 485
PFVCHAN_SYNC_FIELD, 486
PFVFOG_3D_TEX_SIZE, 189
PFVFOG_COLOR, 189
PFVFOG_DENSITY, 189
PFVFOG_DENSITY_BIAS, 189
PFVFOG_EXP, 193
PFVFOG_EXP2, 193
PFVFOG_FLAG_CLOSE_SURFACES, 190
PFVFOG_FLAG_FORCE_2D_TEXTURE, 190
PFVFOG_FLAG_FORCE_PATCHY_PASS, 190
PFVFOG_FLAG_LAYERED_PATCHY_FOG, 190
PFVFOG_LAYERED_MODE, 189
PFVFOG_LIGHT_SHAFT_ATTEN_SCALE, 189
PFVFOG_LIGHT_SHAFT_ATTEN_TRANSLATE,
189
PFVFOG_LIGHT_SHAFT_DARKEN_FACTOR,
189
PFVFOG_LINEAR, 189, 193
PFVFOG_PATCHY_MODE, 189

007-1680-080

Index

PFVFOG_PATCHY_TEXTURE_BOTTOM, 189
PFVFOG_PATCHY_TEXTURE_TOP, 189
PFVFOG_RESOLUTION, 189
PFVFOG_ROTATE_NODE, 189
PFWIN_AUTO_RESIZE, 462
PFWIN_EXIT, 462
PFWIN_GFX_WIN, 463, 474
PFWIN_HAS_OVERLAY, 462
PFWIN_HAS_STATS, 462
PFWIN_NOBORDER, 462
PFWIN_ORIGIN_LL, 462
PFWIN_OVERLAY_WIN, 462, 463, 474
PFWIN_SHARE_GL_CXT, 463
PFWIN_SHARE_GL_OBJS, 463
PFWIN_SHARE_STATE, 463
PFWIN_STATS_WIN, 463, 474
PFWIN_TYPE_NOPORT, 455
PFWIN_TYPE_OVERLAY, 455
PFWIN_TYPE_STATS, 455
toroidal loading, 380
transformations
affine, 615
definied, 756
inheritance through scene graph, 88
order of composition, 616
orthogonal, 615
orthonormal, 616, 623
specified by matrices, 613
transition distance, definition, 756
transparency, 314, 681
traversals
activation, 86
application, 88
attributes, 86
culling, 85, 90, 97
customizing, 90, 97
node pruning, 90
visibility testing, 91-94
database. See databases
definition, 756

007-1680-080

draw, 85, 100-101
intersection, 85, 120-127
TRI format. See formats
Triangle data structure, 549
triangle strip, 551
triangulated irregular networks, 538
trigger routine, definition, 756
tristrip, 551
tsid, 553
values, 552
type system
multiprocessing implications, 734
type, actual, of objects, 17
typographical conventions, xlii

U
UNC format. See formats
up vector, defined, 756
updatable objects, 735
updates, latency-critical, 664
user data, 10
usinit(), 338
usnewlock(), 170, 497, 498
usnewsema(), 497, 498
ussetlock(), 170
usunsetlock(), 170
utilities
cliptexture configuration, 397

V
VClock. See video counter
vector routines, 611

797

Index

vectors
2-component, 611
3-component, 611
4-component, 611
vertex data structure, 555
vertex neighborhoods, 556
vertical stripes, 365
video
counter, 495
dynamic video resolution, 148
field, 637
multiple outputs, 482
output, 482
scan rate, 132
Video Digital Multiplexer (DPLEX), 351
video splitting, 39
view
inset, 38
matrix, 31
offset, 32
view volume visualization, definition, 757
viewing angles, 30
viewing frustum
definition, 757
intersection with, 93
viewing offsets, 32
viewing parameters, 28, 30
viewpoint, 30
definition, 757
viewports, 27
defined, 757
virt_size, 410
virtual addresses and multiprocessing, 734
virtual cliptexture, 382, 429, 446
parameters, 435
set parameters, 431
virtual functions, address space issues, 734
virtual offset, 383

798

virtual reality, xxxix
virtual reality markup language, 89
See also VRML, 89
virtual set, xxxix
virtualLODOffset, 432
VISI bus, 587
visibility culling, 91-94
visual priority. See coplanar geometry
visual simulation, xxxix
visual, defined, 757
VME bus, 591
void, 114
volumes
bounding, 59
boxes, 620
creating, 622
cylinders, 620, 667
dynamic, 59
extending, 623
hierarchical, 91
intersection testing, 624
spheres, 620
visibility testing, 93
boxes, axially aligned, 621
cylinders, 621
geometric, 620
half-spaces, 622
intersections. See intersections
primitives, 620
spheres, 621
transforming, 623
VRML, 89, 258
VRML 2.0, 287

W
Wavefront, 265
wglSwapBuffers(), 674

007-1680-080

Index

widget, defined, 757
windows, 469
WindRiver, 673
WindView, 673
wireframe, 294
workload balancing, 370
WorkShop, 672
write(), 502
wrl format, 287

X
X Windows, 481
XCreateWindow(), 460
XML configuration files, 702, 716
XSGIvc library, 482

Z
z-fighting, 316

007-1680-080

799

Source Exif Data:

File Type                       : PDF
File Type Extension             : pdf
MIME Type                       : application/pdf
PDF Version                     : 1.2
Linearized                      : Yes
Create Date                     : 2002:12:06 10:26:54
Producer                        : Acrobat Distiller 4.0 for Windows
Creator                         : FrameMaker xm5.5P4f
Modify Date                     : 2002:12:06 10:31:21-08:00
Page Count                      : 842
Page Mode                       : UseOutlines

EXIF Metadata provided by EXIF.tools

007 1680 080

Navigation menu

Versions of this User Manual:

Views

Navigation