VISTA Volume Imaging Scalable Toolkit Acronym

1
VISTAVolume Imaging Scalable Toolkit Acronym

• Jon Genetti
• SDSC Scientific Visualization Group

2
Background - Splatter

• Cray T3D/T3E
• In-core (multi-GB)
• Regular, rectilinear data
• Splatting algorithm
• Orthographic projection
• Custom API to AVS module

3
Background - MPIRE

• Massively Parallel Interactive Rendering
  Environment
• Addl platforms IBM SP, Tera, SGI
• Larger in-core (gt10GB)
• Added ray casting
• Added perspective on shared memory
• Added JAVA interface

4
Motivation

• Scalability
• Time-varying data
• Multiple volumes simultaneously
• And ...
• Volume/polygonal data simultaneously
• Orthographic and perspective projections
• Integrate with other visualization toolkits
• Support Alpha Project

5
Scalability

• Larger-than-core imaging
• Handle in-core efficiently
• Big parallel architectures
• Also run on single processor systems

6
Data-order Imaging

• Intelligent data access using ADR
• Prototype shown at SC99
• With Kurc, Chang, Sussman, Saltz (U.Md)

7
Multiple Data Sets

• Efficient use of voxels
• Higher res where needed
• 2GB vs. 40GB
• Level-of-detail

648 x 326 x 648
643
1283
8
Polygonal Data

• Volumes and isosurfaces
• Navigational markers
• Ex stars are billboard polygons with a gaussian
  splat texture map

9
Viewing Models

• Depends on needs of application
• Orthographic
• Faster
• Object viewing
• Perspective
• Less regular data access
• Scene navigation

10
Layered Toolkit Architecture

• Integrates with
• Expression Tree Toolkit
• Mesh Toolkit
• Data Grid Toolkit
• Leverage

Application
VISTA
Expression Tree Toolkit
Mesh Toolkit
Data Grid Toolkit
Data Management
File Format Handling
SRB, ADR, etc.
11
Imaging a Data Grid

• Composite volumes and render from composite

12
Imaging Reduced Expression Tree

• Voxelize sub-trees where appropriate

13
Imaging Full Expression Tree

• Never compute intermediate voxelization
• VISTA evaluates expression tree for each sample
  point needed by renderer
• Best for evolving or time-varying data
• Only resamples where needed