Transcript From Simulation to Visualization: Astrophysics Goes

From Simulation to Visualization:
Astrophysics Goes Hollywood
Frank Summers
January 17, 2002
Simulation
Visualization
Done for research
purposes
Presentation to wide
audience
Simulations
Attempts to replicate & explore aspects of
nature on a computer
 Mathematical abstraction of a physical
process (equations)
 Time sequence

Example: Earth orbiting the Sun
Sun
Earth
Sun
Earth
Sun
Earth
Sun
Earth
Star
Planet
Star
Planet
Object 2
Object 1
Simulation is just numbers …

Position, velocity, and mass
Object Position (AU) Velocity (km/s) Mass (g)
Sun
(0, 0, 0)
(0, 0, 0)
2 x 1033
Earth
(1, 0, 0)
(0, 30, 0)
6 x 1027
… numbers changing over time

Earth position changes
Day
X
Y
Z
1
1.0
0.0
0.0
2
0.9999
0.0172
0.0
3
0.9994
0.0344
0.0
…
…
…
…
364
0.9994
-0.0344
0.0
365
0.9999
-0.0172
0.0
Simulation Details
Initial Conditions
 position, velocity, density, temperature,
etc. for all objects at starting time
 Equations
 gravity, hydrodynamics, radiation,
magnetic fields, expansion of the
universe

Simulation Details
Time Evolution
 Calculate forces, heating, other changes
 Update position, velocity, etc. with new
values
 Repeat
 Data Output
 Write file of positions, velocities, etc
 Series of files covering simulation time

Scientific Accuracy
Simulations expensive, but necessary
 Artist’s conceptions difficult
 Well removed from normal experience
 Complex 3D behaviors
 Coupled feedback between physics
 Scientists can’t describe it sufficiently
 Scientific simulations
 physics equations programmed in
 3D, complexity, and feedbacks included

Visualization

Turn those numbers into pictures
Visualizations
Data Transformation
 Representation
 Choreography
 Rendering
 Compositing

Graphics vs Visualization

Science Graphics
 Pictures, plots, charts
 Illustrations to
scientific argument
 Requires background
knowledge to interpret
 Representational
 Content more
important than form

Scientific Visualization
 Images and movies
 Tells its own story
 Must play off of
audience’s knowledge
 More literal
 Visual message is the
strongest
Data Transformation
Comprehend the dataset
 What quantities?
 What time period?
 What are the assumptions?
 Convert from research quantities to more
generally meaningful quantities

Representation
How literal?
 How artistic?
 How best to promote message?
 How to be least misleading?

Shading
Geometry can get very complex
 ex., surface of an orange
 Shading solution - Use simple shapes
 ex., sphere
 Add complexity when drawing the surface
 Texture - color, pattern
 Bumps - small shape distortions
 Light – reflection, transparency
 Programmability = Flexibility

Shading Example: 3 Balls
Shading Example: Teapot
Shading Stars
Simple Geometry - Disks
Disks w/ Star Shader
Shading Stars
Gaussian
Exponential
Stars by Magnitude
Combination
Calibration using local starfield
Globular
Star Cluster
47 Tucanae
Globular Star Cluster Viz




Data - N-body simulation, 6144 stars (Zwart)
 3D position, absolute brightness, mass, & type
 color derived from mass, type, and spectra*
Stars as point objects
 size depends on apparent brightness
 size calculated in pixels, not 3D space
Star shader in renderman
 calculate app. brightness - each star, each frame
 combination of gaussian and exponential glows
Calibrated by reproducing constellations
Shading Gas Clouds
Choreography
Camera motion
 Invaluable for giving 3D feel
 Missing from most science animations

XY Projection
YZ Projection
XY Projection
YZ Projection
Orion
Nebula
Rendering
Lots of computer time
 Over 2 hours per frame for NASM
 1500 frames – 125 CPU days
 Renderfarms
 Clusters of computers dedicated to
rendering

Renderfarm in 331B




Gathered unused OPO machines
Installed Red Hat Linux
Connected to a switch and created an isolated
private network of five machines
Total out-of-pocket cost $80
Computer 1
Computer 2
Computer 3
Computer 4
Computer 5
Red Hat Linux 7.2
Red Hat Linux 7.2
Red Hat Linux 7.2
Red Hat Linux 7.2
Red Hat Linux 7.2
Dual P3 933 MHz
Dual P3 800 MHz
P2 400 MHz
P2 400 MHz
P3 850 MHz
private network switch
Visualization Wall Schematic
Computing Cluster
Node 1
Node 2
Node 3
Display
Node 4
Node 5
Node 6
1
2
3
4
Node 8
5
9
13
6
10
14
7
11
15
8
12
16
Node 7
Master
Computer
Node 9
Node 10
Node 11
Node 12
Node 13
Node 14
Node 15
Node 16
Compositing
Add multiple elements together
 Provide context

Galaxy Collision Viz

Data - N-body + Hydro simulation
 262,144 particles (Mihos & Hernquist)
 young stars, old stars, gas, dark matter