Conceptos y terminología

Diciembre-2011
Graphics, Virtual Reality and HPC in
the Advanced Computer Graphics Group
Francisco José Serón
Arbeloa
Juan Antonio Magallón
Acknowledge
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
Why
?
Why
?
Period [1980-1984]
• Dr. F. J. Serón made its Ph. D. working in the geophysical
problem of Love earthquake waves propagation. His supervisor
was Dr. J. I. Badal.
• Features
• 2D problem
• Strategy: Modal analysis based on the finite element method
• VAX11-780 [SISD] scalar computer
Why
?
Period [1985-1990]
• Dr. F. J. Serón works with Dr. J. I. Badal in the problem of the
analysis of traveltime and amplitude of full elastic wave
propagation in 2D oil and gas reservoirs.
• Features
• 2D problem
• Strategy: finite element method which produces very big size linear
equations with a huge number of right-hand side vectors
• IBM 3090 VF (Vector Facility) [SIMD] vector computer
Why
?
• Period [1991-2000] Dr. F. J. Sabadell made its PH. D. working
in the problem of the analysis of traveltime and amplitude of
full elastic wave propagation in 2D oil and gas reservoirs. His
supervisors were Dr. F. J. Serón and Dr. J. I. Badal
– Characteristiques of the work
• 2D and 3D problem
• Strategy Based on Domain Decomposition + analysis based on the
finite element method
• Cluster of workstation [MIMD] and Cray T3D
Why
?
Period [1991 - …]
• Dr. F. J. Serón works in the Computer Graphics rendering
problem with different members of the GIGA group, specially
with Dr. J. A. Magallon
– Standard approximation
•
•
•
•
3D problem
Strategy Based on heuristic approximation
of the rendering equation
Real time
GPU´s [SIMD]
– Simulation of the real phenomena
•
•
•
•
3D problem
Strategy Based on Monte-Carlo approximation
of the rendering equation
No real time
Cluster of computers with GPU´s [SIMD + MIMD]
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
Concepts
Computer Graphics
Computer Graphics can be defined as the Information
Technology area using the computer to create, store
and automatically manipulate graphics, real images
and synthetic images.
Virtual Reality
A Virtual Reality system is a paradigm of relationship
between man,
a synthetic world and the computer
that involves real-time simulation and interactions
through multiple sensory channels.
The sensory channels of the human are: sight, hearing, touch, smell and taste.
The ultimate goal is
"Man immersed in an artificial world."
Mixed Reality
A Mixed Reality system is a paradigm of relationship
between man,
their real world and the computer
that involves real-time simulation and interactions
through multiple sensory channels.
The sensory channels of the human are: sight, hearing, touch, smell and taste.
More than immerse a man in an artificial world,
mixed reality
“proposes to enrich his environment."
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
What do you need?
What do you need?
What do you need?
What do you need?
What do you need?
What do you need?
• Peripherals devices
What do you need?
People….
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
Results
Francisco José Serón Arbeloa
Index
Why?
Concepts
What do yo need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
Where are we?
Where are we?
Physically-Based Simulation of Rainbows
Iman Sadeghi, Adolfo Muñoz, Philip Laven, Francisco Seron, Wojciech Jaroz, Diego Gutierrez and
Henrik Wann Jensen
ACM Transactions on Graphics
Where are we?
A Practical Appearance Model for Dynamic Facial Color
Jorge Jimenez, Timothy Scully, Nuno Barbosa, Craig Donner, Xenxo
Alvarez, Teresa Vieira, Paul Matts, Veronica Orvalho, Diego Gutierrez and
Tim Weyrich
Francisco José Serón Arbeloa
ACM Transactions on Graphics, Vol. 29(5) (SIGGRAPH Asia 2010)
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
Hardware used
We used ideas of the HPC world but rarely
Supercomputers
http://www.diegosarmentero.com/2009/05/logo-hpc.html
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
Sample Cases
•
[1996-2001] Centro de Computació i Comunicacions de Catalunya C4
– IBMSP2, SGI PC Array, SGI Origin 2000
•
[1996-1997] Training and Research on Advanced Computing Systems
od Ediburgh (England)
– Cray T3D
Sample Cases
• Problem:
– industrial design (CAF)
– virtual cultural heritage (Sinhaya)
• Features: quality
– ray-tracing, Monte Carlo
– complex scene database
– complex simulation: real lights,
materials, textures…
– multi-channel
video
output
(CAVE, 6 ch), not real time
– separate simluation/visualization
hardware
a)
• System:
– distributed memory HPC cluster
– batch queue system for off-line
render
– 6-node system for visualization
b)
c)
Sample Cases
• Problem:
– industrial lighting simulation
(Indal)
• Features: quality, predictability
– ray-tracing, Monte Carlo
– complex scene database
– complex simulation: real
lights, observer…
– multi-channel video output
(CAVE, 6 ch), not real time
• System:
– distributed system, shared
between
simulation
and
visualization
– simple
in-house
coarse
paralelization
Sample Cases
• Problem:
– inverse reflector design
• Features: optimization problem
– ray-tracing
– search solution in parameter
space
– not-so-hard individual simulation
– many simulations
– failure tolerance
• System:
– distributed memory HPC cluster
– batch queue system (condor)
Sample Cases
• Problem:
– inverse reflector design
• Features: optimization problem
– ray-tracing
– complex simulation, simple
database
– iterative problem, individual
problems not known at start
– each simulation can be adapted
for
distributed
memory
paralelization
(replicated
database)
• System:
– distributed
cluster
– MPI
memory
HPC
Sample Cases
• Problem:
– real time image synthesis
• Features: latency
–
–
–
–
ray-tracing
render each frame in minimal time
complex scene database
user interaction and gui (camera…)
• System:
– multi-core system (Tesla_Nvidia)
– shared memory paralelization:
POSIX threads
Sample Cases
• Problem:
– real
time
synthesis
realistic
image
• Features: latency
– ray-tracing, Monte Carlo
– render each frame in minimal
time
– complex scene database
– Complex simulation: real lights,
soft shadows, fog…
• System:
– multi-core + GPU
– shared memory paralelization:
POSIX threads
– distributed paralelization: CUDA
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
+
• Reality is Real Time.
Francisco José Serón
2011
Reality but not real time
Francisco José Serón Arbeloa
Francisco J. Seron, Enrique Meléndez Estrada
Ecuación FRE (Full Radiance Equation)
emisividad
fosforescencia
Interreflexión + fluorescencia
emisividad
fosforescencia
Medio no participativo
Medio participativo
Interreflexión + fluorescencia
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
What do we need?
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
Evolution
From 3 millions to 3.000 millions
of transistors in 15 years
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
What do we convey?
• We know, the growing trend of the use of generalpurpose computing on GPUs (GPGPU) may lead to
wider use of SIMD and MIMD in the future of the
computer graphics simulation phenomena.
– [SIMD] In total there is 180 millions de Cuda Gpu´s and
more of 100.000 developers.
– [MIMD] In total 17 of the Top500 supercomputers use
GPUs as accelerators. Six of those use the Cell processor,
ten use Nvidia chips and one uses ATI Radeon chips.
The predictable future
The predictable future
“I am not sure
how I will program a Petaflop computer,
but I am sure that I will need MPI somewhere”
– HDS 2001
Index
Why?
Concepts
What do you need?
Our results
Where are we?
Hardware used
Sample Cases
What is reality?
What do we need?
The Evolution of Computer Graphics
What do we convey
Conclussion
Conclussion
Francisco José Serón Arbeloa
Thanks