Docstoc

Isosurfaces-on-Display-Wall

Document Sample
Isosurfaces-on-Display-Wall Powered By Docstoc
					  Fast Isosurface Visualization on a
High-Resolution Scalable Display Wall

          Adam Finkelstein
            Allison Klein
                Kai Li

         Princeton University

      Sponsors: DOE, Intel, NSF
Overview


The display wall environment
   Motivation

   Challenges



Isosurfaces on the display wall
    Extraction

    Rendering

    Future directions
Some snapshots
Some snapshots
Some snapshots
Some snapshots
      Scalable low-cost display wall



    128-CPU                  Commodity
Production cluster            projectors




 64-CPU cluster
                       Extensible
                        Router

32-node I/O cluster                  PCs w/                ...
                                     3-D accelerators
      Wireless links
             PPPL LAN vBNS                        T1
                                                        AT&T Lab
Technology trends




                    Time
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
Multi-projector displays

SGI-based displays
   Government labs:
    ANL, LANL, LLNL, Sandia
   Industry:
    AT&T, Panoram Tech, Trimension, ...
   Universities:
    Minnesota, Stanford, UI Chicago, UNC
PC-based displays
   Princeton, Intel, ANL

   Next: Illinois, LLNL, Sandia, Lucent, ...
First Video
Research challenges

   Parallel rendering
   Fast communication
   Seamless imaging
   Interaction techniques
   Spatialized sound
   Virtual environments
   Visualization systems
Visualization of isosurfaces
Goals

Large data sets
    Visible woman

    Astrophysical simulations

Large display
    Inexpensive

    High resolution

Interactive rates
    Extraction

    Rendering
Runtime components

Extraction
  Find voxels containing the isosurface.

Communication
  Send surface information to display.

Rendering
  Draw the surface.
    Runtime architecture

Extraction     Communication      Rendering



                  network
 database                      display
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]
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]
Filtering search

0.00 
0.12 
0.38 
0.57 
0.61 
0.78 
0.93 
Benefits of filtering search

   Nice space / time tradeoff
   Better asymptotic worst case
   Very easy to code
   Trivially parallelizeable
    Runtime architecture

Extraction     Communication      Rendering



                  network
 database                      display
    Runtime architecture

Extraction     Communication      Rendering



                  network
 database                      display
Communication

Gigabit network (Myrinet)
Scalable
Virtual memory mapped communication
Currently we ship voxels:
    voxel ID

    marching cube case

    edge interpolants
    Runtime architecture

Extraction     Communication      Rendering



                  network
 database                      display
Rendering

Rely on PC graphics cards
Static screen-space partitioning
Current bottleneck
Edge blending
Second Video
How do we make it faster?

Rendering:
   Next generation of graphics cards

   Load balancing



General:
   Surface simplification

   Multiresolution representations
Broader directions

Other vis techniques
Remote visualization
   Compression
   Networking: PPPL, AT&T, CorridorOne

Scalable storage server
   3 TB storage
   1.5 GB / sec

   Intelligent caching

   $150K

				
DOCUMENT INFO
Shared By:
Tags: Isosu, rface
Stats:
views:5
posted:12/1/2009
language:English
pages:29
Description: Isosurfaces-on-Display-Wall