9th International Fall Workshop VISION, MODELING, AND VISUALIZATION 2004 November 16 - 18, 2004 Stanford (California), USA Poster Session 3 Simulation.

9th International Fall WorkshopVISION, MODELING, AND VISUALIZATION 2004November 16 - 18, 2004Stanford (California), USA

Poster Session 3

Simulation and Rendering

Arno Zinke, Gerrit Sobotka, Andreas Weber


Photo Realistic Rendering Of Blond Hair

Institute of Computer Science II, University of Bonn, Germany

Photo Realistic Rendering Of Blond Hair

indirect illumination close upshair-hair scattering

“near field“ scattering model for light from hair fibers generalization of Marschner et al. 2003 (“far field” model)

foto near field

far field

&complex lighting

far

fiel

d m

od

el

front light back light

nea

r fi

eld

mo

de

l

Nils Thürey, Ulrich Rüde


Free Surface Lattice-Boltzmann fluid simulations with and without level sets

University of Erlangen, LSS

Free Surface Fluid Simulations

• Lattice-Boltzmann methods for free surfaces

• Comparison of two surface tracking algorithms

• Visualization results (movies!)

Veronica Sundstedt

Alan Chalmers, Kirsten Cater, Kurt Debattista


Top-Down Visual Attention for Efficient Rendering of Task

Related Scenes

Department of Computer Science University of Bristol, UK

Top-Down Visual...• People fail to notice quality degradations

in images when performing a visual task

• We exploit this in a perception-based selective rendering framework

14.5 min 14.5 min 3 min


Reducing State Changeswith a Pipeline Buffer

Jens Krokowski, Christian Sohler

University of PaderbornGermany

Matthias Westermann

Dortmund University

Germany

Harald Räcke

Carnegie Mellon UniversityUSA

Reducing State Changeswith a Pipeline Buffer

• Problem: state sorting vs. spatial sorting

• Our new method:– Application can use any sorting – Online state sorting during rendering

• Main results:– Reduces # state changes significantly– Improves the rendering time ~30%– Achieves almost the same rendering

time as a presorted sequence

Selection Strategy

Application

OD

-Buffe

r

State TrackingPipeline Buffer

SIS=SG

H

Graphics Hardware

S. Kimmerle, M. Wacker, C. Holzer


Multilayered Wrinkle Textures from Strain

WSI/GRIS University Tübingen

Multilayered Wrinkle Texturesfrom Strain

Enrich simulated garments with textures to get finer details

• Based on strain deformation tensor• Procedurally generated• Blend different textures depending on strain

directions• Apply textures by bump and displacement

mapping• Much faster than simulation of high-resolution

meshes• Physical plausible results

Ugo Erra, Rosario De Chiara,Vittorio Scarano, Maurizio Tatafiore


Massive Simulation using GPU of a distributed behavioral model of a flock with obstacle avoidance

ISIS Lab.Dipartimento di Informatica ed Applicazioni “R.Capocelli”

Università di Salerno

Massive Simulation using GPU of a distributed behavioral model of a flock with obstacle avoidance

• Goals– Map a behavioral model of a flock on GPU to manage efficiently large

aggregate motion of “birdoids”(flock of birds, herd of land animals, bank of fishes).

– Use graphic device extensions for obstacle avoidance of the flock.• Tools

– Birdoids are managed as pixels of rectangular texture.– 3D texture and hardware interpolation are used as look-up tables to

manage vector fields.– Pixel shaders use the behavioral model to update the position of

birdoids. • Results

– GPU boosted simulation scales well on increasing number of boids– On GeForce FX 5800(average FPS)

• 32 birdoids - 115 FPS• 4096 birdoids - 30 FPS• 16384 birdoids - 8 FPS

N. Fritz, Ph. Lucas, Ph. Slusallek


CGiS, a new Language for Data-Parallel GPU Programming

Saarland University

CGiS, a new Language for Data-Parallel GPU Programming

• Trend: GPGPU• Problem: Abstraction

from hardware• Main obstacle: Multiple

domains• Solution: CGiS

– Unified language– Abstract– Robust– Portable

• Feedback from (potential) users wanted

MainApplication

SSEKernels

GPUKernels

GPU-RT SSE-RT

CGiSSource

OptimizingCGiS

Compiler

GlueCode

Kirill Dmitriev and Hans-Peter Seidel


Progressive Path Tracing with Lightweight Local Error

Estimation

Max-Planck-Institut für Informatik,Saarbrücken (Germany)

Lightweight error estimation

• Split all samples to two equal halves A and B• Estimate accuracy for 4x4 pixels tile i as:

Reference image Estimated error Number of samples

Manfred ErnstChristian Vogelgsang

Günther Greiner


Stack Implementation onProgrammable Graphics Hardware

Universität Erlangen-Nürnberg

Lehrstuhl für Graphische Datenverarbeitung

Stack Implementation on Programmable Graphics Hardware

• Essential for hierarchical data structures

• Stacks are stored in textures

• Push and pop with fragment programs

• Multiple stacks

• GPU ray tracing using Kd-trees

Martin Sunkel, Jan Kautz, H.-P. Seidel


Rendering and Simulation of Liquid Foams

Max Planck Institute for Computer Science

Rendering and Simulation of Liquid Foams

Two-stages:

• Physics on CPU– Sphere interaction– Intersection planes

• Geometry computation on GPU– Project intersecting vertices to planes

Marco Winter


Depth-Buffer based Navigation

University of Erlangen-Nuremberg

Depth-Buffer based Navigation

• Navigation system for avatars

• Collision handling solely based on depth-buffer analysis

• Fast, intuitive

9th International Fall Workshop VISION, MODELING, AND VISUALIZATION 2004 November 16 - 18, 2004 Stanford (California), USA Poster Session 3 Simulation.

Documents