CS 551/851 Big Data in Computer Graphics Greg Humphreys

advertisement
CS 551/851
Big Data in Computer Graphics
Greg Humphreys
What does “big” mean?
“I cannot define it, but I know it when I see it”
- Justice Potter Stewart
• “Big” is a relative term
• It happens whenever a resource is fully
consumed
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Models
Pratt-Whitney 6000 turbine engine and rotor blade
120 million cell calculation, 500,000 triangle surface
Stanford Center for Integrated Turbulence Simulations
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Models
Double Eagle Tanker Model: 83 million triangles
UNC Walkthrough Project
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Models
Scans of Saint Matthew (386 MPolys) and the David (2 GPolys)
Stanford Digital Michelangelo Project
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Displays
Window system and large-screen interaction metaphors
François Guimbretière, Stanford University HCI group
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Displays
Simulation of Compressible Turbulence (2K x 2K x 2K mesh)
Sean Ahern and Randall Frank, LLNL
Big Data in Computer Graphics Fall 2002 Lecture 1
Big LCD Displays
2400
3840
Jet engine nacelle model courtesy Goodrich Aerostructures
Peter Kirchner and Jim Klosowski, IBM T.J. Watson
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Sloppy Displays
WireGL extensions for casually aligned displays
UNC PixelFlex team and Michael Brown, UKY
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Texture Maps
153K x 153K = 73GB!!
Using Texture Mapping with Mipmapping to Render a VLSI Layout
Solomon and Horowitz, DAC 2001
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Dynamic Range
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Dynamic Range
1/1000
1/500
1/250
1/125
1/60
1/30
1/15
1/8
1/4
Gradient Domain High Dynamic Range Compression
Fattal, Lischinski and Werman, SIGGRAPH 2002
Big Data in Computer Graphics Fall 2002 Lecture 1
Big Chips
•
•
•
•
•
•
•
GeForce4 die plot courtesy NVIDIA
Big Data in Computer Graphics Fall 2002 Lecture 1
63 MTransistors
1.23 TOps/sec (!)
10 GB/sec
136 MTris/sec 
1.2 GPix/sec
4 rendering pipes
8 textures
Big… Everything
Realistic Modeling and Rendering of Plant Ecosystems
Deussen, Hanrahan, Lintermann, Mech, Pharr and Prusinkiewicz, SIGGRAPH 1998
Big Data in Computer Graphics Fall 2002 Lecture 1
What Once Was Big…
1 mo.
106 s
log time
Courtesy Frank Crow, Interval
Fanatical
1 week
1 day
104 s
Possible Teddy Bear 250 GI’s
1 hr.
100 s
Practical
1 min.
Kitchen Table 10 GI’s
Interactive
1.0 s
10 gips
1 gips
100 mips
10 mips
Big Data in Computer Graphics Fall 2002 Lecture 1
100 gips
Immersive
0.01 s
Stemware 100 MI’s
log performance
Slide courtesy Pat Hanrahan and Kurt Akeley
Course Information
•
•
•
•
•
•
Seminar-style: Read + discuss
Tuesday/Thursday 2:00-3:15 in Olsson 228E
Office hours MW 10:00-12:00 in Olsson 216
Discussions will be student-led
One assignment, one project
Course web page:
http://www.cs/~gfx/Courses/2002/BigData
• This is an experiment. Feedback is crucial!
Big Data in Computer Graphics Fall 2002 Lecture 1
Discussions
• Each student will lead at least one class
• Prepared presentation for 30-45 minutes:
–
–
–
–
Background information
Paper summaries
Key ideas
Interruptions encouraged
• Guide discussion
• All students will submit 2-3 questions about
the reading before class, use those as a
starting point
• Starting 9/10 (I’ll do the first three)
Big Data in Computer Graphics Fall 2002 Lecture 1
Assignment 0
• Choose days to present
• Submit your first three choices
• Due evening of 9/3
Big Data in Computer Graphics Fall 2002 Lecture 1
Assignment: Benchmarking
• Probe performance characteristics of
graphics hardware
• Basics: triangle/fill rates, texture download
• Extras
–
–
–
–
–
Triangle areas/shapes
Texture cache
Vertex cache
Interface bottleneck
Others?
• Due September 26th
Big Data in Computer Graphics Fall 2002 Lecture 1
Projects
• Two months investigating something cool
• Need not be novel, but it helps (especially
for you graduate students)
• Can work in groups no larger than 2
• Writeup quality important: treat it as a
conference submission
• Topic proposal due October 3rd
• Writeup/presentations due December 3rd
• Consider publishing your work…
Big Data in Computer Graphics Fall 2002 Lecture 1
About Greg
• B.S.E. Princeton, 1997
• Ph.D. Stanford, 2002
• CTO, Ahpah Software
(Reverse-engineering
technology)
• Research focus on
scalable rendering
using commodity technology: “Chromium”
• Writing textbook on Image Synthesis (class next
semester)
• Looking for students who like serious hacking (hint)
Big Data in Computer Graphics Fall 2002 Lecture 1
Download