1. GPU-based Raycasting of Volumetric Data
GPU-based Raycasting of Volumetric Data
Dongliang Chu
SIParCS-2015
CISL/HSS/VAPOR

2. GPU-based Raycasting of Volumetric Data
Agenda
 Raycasting
 GPU-based Raycasting
 Raycasting In Layered Data
 Proposed Adaptive-Sampling Raycasting
 Accomplishment
 Conclusion
 Future work
2

3. GPU-based Raycasting of Volumetric Data
Raycasting
 Raycasting in scientific visualization
3
Isosurface Extraction Direct Volume Rendering
Volumetric Data for Simulation of Turbulence

4. GPU-based Raycasting of Volumetric Data
GPU-based Raycasting
 Rendering Pipeline on GPU
4
Geometry
Processing &
Rasterization
Fragment
Processing
Vertex
Processing
Vertex
Data
Primitive
Data
Frame
Buffer
Vertex
Shader
Fragment
Shader
Raycasting
….

5. GPU-based Raycasting of Volumetric Data
Raycasting for Volume Rendering
 Algorithm animations
5

6. GPU-based Raycasting of Volumetric Data
Raycasting for Isosurface extraction in
Layered Grid and Data
 Uniform Grid vs Layered Grid
6
(0,0) 0.2 0.4 0.6 0.8 1.0
0.2
0.4
0.6
0.8
1.0
(0,0) 0.2 0.4 0.8 1.0
0.1
0.4
0.5
0.9
1.0
0.6X
Y
X
Y
0
20
80
100
20
0
0
20
80
100
20
0
DensityValue
DensityValue
, Layered Data

7. GPU-based Raycasting of Volumetric Data
Raycasting In Layered Grid and Data
 Practical layered ocean data
7
(0,0) 0.2 0.4 0.8 1.0
-0.05
-0.1
-0.15
-0.7
0.6
X
Y
0
20
80
100
20
0 DensityValue
-0.2

8. GPU-based Raycasting of Volumetric Data
Raycasting In Layered Grid and Data
 Real-life illustrating example
8
Isosurface of salinity from data generated by a regional ocean model
64 samples 1024 samples

9. GPU-based Raycasting of Volumetric Data
Proposed Adaptive-Sampling
Raycasting
 Adaptive-Sampling
9
(0,0) 0.2 0.4 0.8 1.0
0.2
0.4
0.6
0.8
1.0
0.6 X
Y
0
20
80
100
20
0
FieldValue

10. GPU-based Raycasting of Volumetric Data
Proposed Adaptive-Sampling
Raycasting
 Algorithm brief description
1) Given the very first enter point corrresponding to first cell
2) Calculate the corresponding exit point: calculating the ray’s
possible intersecting points with three hiting- possible
boundary planes, take the point closest to enter pointer as the
exit point, set next cell to which the hit boundary plane
belong, but not covered yet
3) turn the exit point into the enter point for next cell, goto step
2)
1
0

11. GPU-based Raycasting of Volumetric Data
Proposed Adaptive-Sampling
Raycasting
 Difficulties
 No Debugger
 No “printout” statement
 Numeric calculation inaccuracy
1
1

12. GPU-based Raycasting of Volumetric Data
Proposed Adaptive-Sampling
Raycasting
 Solution
 Use each pixel’s color channel to for status
checking
 Countless loops of analysing, assumption, test,
proof/disproof
1
2

13. GPU-based Raycasting of Volumetric Data
Accomplishment
 Test data
1
3
64^3 volumetric data with 5 non-zero valued layers, 4 condensed, the other apart

14. GPU-based Raycasting of Volumetric Data
Accomplishment
 Running Result (1)
1
4
Proposed Algorithm Original Algorithm

15. GPU-based Raycasting of Volumetric Data
Accomplishment
 Running Result (2)
1
5
Proposed Algorithm Original Algorithm

16. GPU-based Raycasting of Volumetric Data
Accomplishment
 Running Result (3)
1
6
Proposed Algorithm Original Algorithm

17. GPU-based Raycasting of Volumetric Data
Conclusion
 GPU Programming is big challenge
 Implement a self-adaptive raycaster working
well with ocean and atmosphere data
1
7

18. GPU-based Raycasting of Volumetric Data
Future Work
 Apply the proposed algorithm to curvilinear
data for isosurface extraction
 Apply the proposed algorithm for direct
volume rendering
1
8

19. GPU-based Raycasting of Volumetric Data
1
9

20. GPU-based Raycasting of Volumetric Data
Open Discussion
 Programming Experience
• Tips for improving
• Lessons for warning
2
0