Fast Isosurface Visualization on a HighResolution Scalable Display Wall

1
Fast Isosurface Visualization on
aHigh-Resolution Scalable Display Wall

• Adam Finkelstein
• Allison Klein
• Kai Li
• Princeton University
• Sponsors DOE, Intel, NSF

2
Overview

• The display wall environment
• Motivation
• Challenges
• Isosurfaces on the display wall
• Extraction
• Rendering
• Future directions

3
Some snapshots
4
Some snapshots
5
Some snapshots
6
Some snapshots
7
Scalable low-cost display wall
128-CPU Production cluster
Commodity projectors
64-CPU cluster
Extensible Router
PCs w/ 3-D accelerators
32-node I/O cluster
...
Wireless links
T1
PPPL
LAN
vBNS
ATT Lab
8
Technology trends
9
Scalable low-cost display wall

• Now
• 8 18 rear-projection screen
• 8 polysilicon LCD projectors deliver 6 million
  pixels per frame (4096 x 1536)
• A network (Myrinet) of 15 Pentium-II 450Mhz (8
  have Intergraph graphics accelerators)
• Soon
• 15 new-generation projectors will deliver 20
  million pixels per frame (6400 x 3072)
• A network of new-generation PCs with
  new-generation 3D graphics accelerators

10
Multi-projector displays

• SGI-based displays
• Government labsANL, LANL, LLNL, Sandia
• Industry ATT, Panoram Tech, Trimension, ...
• UniversitiesMinnesota, Stanford, UI Chicago,
  UNC
• PC-based displays
• Princeton, Intel, ANL
• Next Illinois, LLNL, Sandia, Lucent, ...

11
First Video
12
Research challenges

• Parallel rendering
• Fast communication
• Seamless imaging
• Interaction techniques
• Spatialized sound
• Virtual environments
• Visualization systems

13
Visualization of isosurfaces
14
Goals

• Large data sets
• Visible woman
• Astrophysical simulations
• Large display
• Inexpensive
• High resolution
• Interactive rates
• Extraction
• Rendering

15
Runtime components

• Extraction
• Find voxels containing the isosurface.
• Communication
• Send surface information to display.
• Rendering
• Draw the surface.

16
Runtime architecture
Extraction
Communication
Rendering
network
17
Extraction on one processor

• Acceleration methods Cignoni97
• Spatial -- e.g. octree Parker,Shen
• Seed -- e.g. seed and traverse Bajaj
• Value -- e.g. interval tree Cignoni

18
Extraction on one processor

• Acceleration methods Cignoni97
• Spatial -- e.g. octree Shen
• Seed -- e.g. seed and traverse Bajaj
• Value -- e.g. interval tree Cignoni
• We use filtering search Chazelle86

19
Filtering search

• 0.00 ?
• 0.12 ?
• 0.38 ?
• 0.57 ?
• 0.61 ?
• 0.78 ?
• 0.93 ?

20
Benefits of filtering search

• Nice space / time tradeoff
• Better asymptotic worst case
• Very easy to code
• Trivially parallelizeable

21
Runtime architecture
Extraction
Communication
Rendering
network
22
Runtime architecture
Extraction
Communication
Rendering
network
23
Communication

• Gigabit network (Myrinet)
• Scalable
• Virtual memory mapped communication
• Currently we ship voxels
• voxel ID
• marching cube case
• edge interpolants

24
Runtime architecture
Extraction
Communication
Rendering
network
25
Rendering

• Rely on PC graphics cards
• Static screen-space partitioning
• Current bottleneck
• Edge blending

26
Second Video
27
How do we make it faster?

• Rendering
• Next generation of graphics cards
• Load balancing
• General
• Surface simplification
• Multiresolution representations

28
Broader directions

• Other vis techniques
• Remote visualization
• Compression
• Networking PPPL, ATT, CorridorOne
• Scalable storage server
• 3 TB storage
• 1.5 GB / sec
• Intelligent caching
• 150K

29
(No Transcript)