adaptive visualization of dynamic unstructured meshesstevec/slides/proposal/proposal.pdfthe proposal...
TRANSCRIPT
Adaptive Visualization of Dynamic Unstructured
Meshes
Steven P. Callahan
Advisor: Dr. Cláudio T. Silva
Scientific Computing and Imaging Institute,University of Utah
The Problem
‣ Large-scale simulations produce a lot of data
‣ Interactive visualization techniques not keeping up
2
The Problem
Earthquake Simulation14 Million Tetrahedra
CORIE EOFS6 Million Tetrahedra
96 Scalar Time Steps
Heart Simulation48 Higher Order Cells920 Vertex Time Steps
3
The Vision
‣ Data exploration
‣ Fast interaction
‣ Multiple configurations
4
The Vision
‣ Data exploration
‣ Fast interaction
‣ Multiple configurations
5
The Vision
‣ Data exploration
‣ Fast interaction
‣ Multiple configurations
6
The Proposal
‣ Adaptive Visualization of Dynamic Unstructured Meshes
‣ Adaptive Volume Rendering
‣ Dynamic Scalars
‣ Dynamic Geometry
7
Adaptive Volume Rendering
‣ Hardware-Assisted Visibility Sorting (HAVS)
‣ Sort in Object Space (CPU)
‣ Sort in Image Space (GPU)
‣ http://havs.sf.net
‣ 130 downloads (~15/mo.)
‣ VTK/ParaView
‣ Hardware Proposal
Unstructured Volume Rendering
Algorithms vs. Hardware
1
10
100
1000
10000
100000
1000000
10000000
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Te
tra
he
dra
/Se
co
nd
Algorithms
Hardware
Projected
Tetrahedra
Incremental
Slicing
GATOR
HW Ray
Casting
HAVS
Log Scale
8
Adaptive Volume Rendering
‣ Dynamic Level-of-Detail
‣ Domain- vs. Sample-based simplification
‣ Importance sampling
‣ Dynamically adjusts the number of triangles rendered at each frame
‣ 10-15 million tets/sec
r r r
t
g1(t) g
2(t)g(t)
tt
9
Adaptive Volume Rendering
‣ Point-Based Volume Rendering
‣ Points are more flexible
‣ Large datasets have subpixel-size geometry
‣ Minimize error by approximating cells
‣ LOD improved with point resizing
10
Progressive Volume Rendering
3%0.01 sec
33%7 sec
66%18 sec
100%34 sec
11
Progressive Rendering
‣ Progressive Rendering‣ Show intermediate results‣ Reuse intermediate results‣ Allow user interrupt‣ Only render pertinent data
‣ Client-Server Architecture‣ Support a thin client with limited memory‣ Standard server used as a data repository‣ Facilitate remote visualization
12
Progressive Rendering
‣ The Server
‣ Preprocess
‣ Geometry Server
‣ Octree Traversal
‣ Object-Space Sort
13
Progressive Rendering
‣ The Client
‣ Preprocess
‣ Interactive Mode
‣ Boundaries only
‣ Progressive Mode
‣ Complete, Active, and Progressive buffers
‣ Completed Mode
‣ Final image displayed and stored
ProgressiveComplete Active
14
Progressive Rendering
‣ Facilitates fine-scale exploration
‣ Allows remote and out-of-core rendering for large datasets
‣ A variety of configurations are possible
15
Time-Varying Scalar Fields
‣ Volume Rendering
‣ Dynamic Level-of-Detail
‣ Compression & Data Transfer
‣ Parallel Processing16
Time-Varying Scalar Fields
‣ Volume Rendering
‣ Dynamic Level-of-Detail
‣ Compression & Data Transfer
‣ Parallel Processing17
Time-Varying Scalar Fields
‣ Volume Rendering
‣ Dynamic Level-of-Detail
‣ Compression & Data Transfer
‣ Parallel Processing18
Time-Varying Scalar Fields
‣ Importance sampling for time varying scalar fields
‣ Local sampling
‣ Global sampling
19
Time-Varying Scalar Fields
‣ Volume Rendering
‣ Dynamic Level-of-Detail
‣ Compression & Data Transfer
‣ Parallel Processing20
Time-Varying Scalar Fields
Uncompressed Compressed21
Time-Varying Scalar Fields
‣ Volume Rendering
‣ Dynamic Level-of-Detail
‣ Compression & Data Transfer
‣ Parallel Processing22
Time-Varying Scalar Fields
‣ Achieves about 3 Million tetrahedra/second
‣ Only a 5% increase over static rendering
‣ Adaptive for large datasets
‣ Efficient use of resources
23
Time-Varying Geometry
‣ Future Work
‣ Focus on compression/decompression
‣ Data transfer may need to be revisited
‣ Point-based or triangle based rendering?
24
‣ Dynamic Topology
‣ Higher Order Elements
‣ Future Hardware
Out of Scope
25
The Progress
Adaptive Volume Rendering
Time-Varying Scalars
Time-Varying Geometry
26
The Progress
‣Hardware-Assisted Visibility Sorting for Unstructured Volume Rendering, S. Callahan, M. Ikits, J. Comba, and C. Silva, IEEE Transactions on Visualization and Computer Graphics, 11(3):285–295, 2005.‣Interactive Rendering of Large Unstructured Grids Using Dynamic Level-Of-Detail, S. Callahan, J. Comba, P. Shirley, and C. Silva. IEEE Visualization, pp. 199–206, 2005. ‣Progressive Volume Rendering of Large Unstructured Grids, S. Callahan, L. Bavoil, V. Pascucci, C. Silva, IEEE Transactions on Visualization and Computer Graphics (Proceedings of Visualization 2006), 12(5), pp. 1307-1314, 2006. ‣Interactive Volume Rendering of Unstructured Grids with Time-Varying Scalar Fields, F. Bernardon, S. Callahan, J. Comba, and C. Silva. Eurographics Symposium on Parallel Graphics and Visualization, pp. 51–58, 2006.. ‣An Adaptive Framework for Visualizing Unstructured Grids with Time-Varying Scalar Fields, F. Bernardon, S. Callahan, J. Comba, C. Silva. Parallel Computing, 2006, submitted.‣Multi-Fragment Effects on the GPU using the k-Buffer, L. Bavoil, S. Callahan, A. Lefohn, J. Comba, C. Silva, ACM Symp. on Interactive 3D Graphics and Games (i3D). 2007, to appear‣Hardware-Assisted Point-Based Volume Rendering, E. Anderson, S. Callahan, C. Scheidegger, J. Schreiner, C. Silva. 2006, submitted.‣iRun: Interactive Rendering of Large Unstructured Grids, H. Vo, S. Callahan, N. Smith, C. Silva, W. Martin, D. Owen, D. Weinstein. 2006, submitted.‣Interactive Transfer Function Specification for Direct Volume Rendering of Disparate Volumes, F. Bernardon, L. Ha, S. Callahan, J. Comba, C. Silva. 2006, submitted.
27
Dissertation Outline
1 Introduction2 Related Work3 Background
3.1 HAVS3.2 Dynamic LOD
4 Adaptive Volume Rendering4.1 Point-based Volume Rendering4.2 Progressive Volume Rendering
5 Time-Varying Scalar Fields6 Time-Varying Geometry7 Discussion8 Conclusion
Target Graduation: Spring 2008Target Employment: Academia/Research Institution
28
Acknowledgments
‣ Advisor‣ Cláudio T. Silva
‣ Collaborators‣ Erik W. Anderson‣ Louis Bavoil‣ Fábio F. Bernardon‣ João L. D. Comba‣ Milan Ikits‣ Linh H. Ka‣ Aaron Lefohn‣ Valerio Pascucci‣ John Schreiner ‣ Peter Shirley‣ Huy T. Vo
29