Gamasutra

Rendering & Simulation with GPUs

Wednesday, 27 July, 3:45 pm - 5:55 pm, Anaheim Convention Center, Ballroom C
Session Chair: Oliver Wang, Adobe Systems, Inc.

Rendering on A Power Budget: A Power-Optimal Rendering Framework

This paper presents a real-time, power-optimal rendering framework that finds the optimal rendering settings to minimize power consumption while maximizing visual quality. The method includes a novel power-error, multi-objective cost space, and formally formulates power saving as an optimization problem.

Rui Wang
Zhejiang University

Bowen Yu
Zhejiang University

Julio Marco
Universidad de Zaragoza

Tianlei Hu
Zhejiang University

Diego Gutierrez
Universidad de Zaragoza

Hujun Bao
Zhejiang University

A System for Rapid Exploration of Shader Optimization Choices

This shading language and compiler framework facilitate rapid exploration of shader-optimization choices. The abstraction extends the scope of shader execution beyond traditional GPU graphics pipelines and enables a diverse set of shader optimizations to be described by a single mechanism.

Yong He
Carnegie Mellon University

Tim Foley
NVIDIA Corporation

Kayvon Fatahalian
Carnegie Mellon University

Efficient GPU Rendering of Subdivision Surfaces Using Adaptive Quadtrees

This goal of this novel method for real-time rendering of subdivision surfaces is to make subdivision faces as easy to render as triangles, points, or lines. The approach uses the GPU tessellation hardware and processes each face of a base mesh independently and in a streaming fashion.

Wade Brainerd
Activision Blizzard, Inc.

Tim Foley
NVIDIA Corporation

Manuel Kraemer
NVIDIA Corporation

Henry Moreton
NVIDIA Corporation

Matthias Nießner
Stanford University

Ebb: A DSL for Physical Simulation on CPUs and GPUs

Ebb is a performance-portable domain-specific language (DSL) for writing physical simulations. Ebb shows how high-performance DSLs can support more complex data models.

Gilbert Bernstein
Stanford University

Chinmayee Shah
Stanford University

Crystal Lemire
Stanford University

Zachery DeVito
Stanford University

Matthew Fisher
Stanford University

Philip Levis
Stanford University

Pat Hanrahan
Stanford University

Simit: A Language for Physical Simulation

Simit is a new high-performance language for writing simulations. Simit programs are typically shorter than Matlab programs, but they are competitive with hand-optimized code, and they also run on GPUs.

Fredrik Kjolstad
Massachusetts Institute of Technology

Shoaib Kamil
Adobe Systems Incorporated

Jonathan Ragan-Kelley
Stanford University

David I.W. Levin
Disney Research

Shinjiro Sueda
California Polytechnic State University

Desai Chen
Massachusetts Institute of Technology

Etienne Vouga
University of Texas at Austin

Danny M. Kaufman
Adobe Systems Incorporated

Gurtej Kanwar
Massachusetts Institute of Technology

Wojciech Matusik
Massachusetts Institute of Technology

Saman Amarasinghe
Massachusetts Institute of Technology

Parallel Inverse Kinematics for Multithreaded Architectures

This paper solves damped least-squares inverse kinematics using a parallel line search by sampling three critical input parameters. The algorithm can handle complex articulated bodies - up to 600 degrees of freedom - at interactive frame rates. Implementations are 10 to 150 times faster compared to a state-of-the-art serial implementation.

Pawan Harish
École polytechnique fédérale de Lausanne

Mentar Mahmudi
École polytechnique fédérale de Lausanne

Benoît Le Callennec
Moka Studio

Ronan Boulic
École polytechnique fédérale de Lausanne