CUDA: Week in Review
Wed, Nov. 10, 2010, Issue #41 - Newsletter Home
WELCOME
Welcome to CUDA: Week in Review, an online news summary for the worldwide CUDA and GPU computing community.
CONTENTS
CUDA SPOTLIGHT
At the Intersection of Art and Technology
Visual computing artist Philipp Drieger of Eichstatt, Germany, used CUDA to build a "Two Million Pixel Experiment" exploring a computational approach to art. In the experiment, an original video created by Philipp was mapped into 3D. Here's an extract of our interview:

NVIDIA: Philipp, what inspired you to do this experiment?
Philipp: As an artist, my deeper inspiration was solely aesthetic. As a visual computing enthusiast, I started with this question: "Is it possible to map an HD video source of two megapixels in 1080p into a 3D space - frame by frame, pixel by pixel - in real-time? "
 
NVIDIA: What were the challenges?
Philipp: It's not easy to calculate two million pixels! And, it gets much harder if these pixels are mapped into 3D. Each vertex has to be scaled by the luminance value of the corresponding pixel, which means many millions of calculations per second have to be done.
 
NVIDIA: What role did CUDA play in your experiment?
Philipp: Each frame is processed in real-time on the GPU using CUDA. This experiment could not have been done without CUDA and the parallel computing capabilities of NVIDIA GPUs.
 
NVIDIA: What did you learn from it?
Philipp: From a programming perspective, I learned how to create a CUDA application that uses the following libraries in a successful interaction:
1. DirectShow.NET for grabbing single frames of a video source
2. CUDA.NET to access CUDA technology in C#/.NET
3. SlimDX for presenting the results in a DirectX 11 context
 
NVIDIA: Are you working on other projects using CUDA?
Philipp: I recently implemented CUDA for laying out large-scale graph structures in 3D. This provided dramatic speed-ups, even for very complex graphs. I'm currently developing an 'interactive visualization system' that will leverage CUDA for performance optimizations.
 
NVIDIA: Tell us about your company, noumentalia.de.
Philipp: noumentalia.de is a digital arts company that offers IT services, programming and consulting as well as fine art content creation. Visualcompute.com is a new domain that we set up to showcase the intersection between art and technology, especially GPU-driven real-time applications using CUDA.

To learn more about Philipp's work, contact him at info@visualcompute.com or visit www.noumentalia.de or www.visualcompute.com. To see the YouTube video, go to: http://www.youtube.com/watch?v=kHhkLyJLLYI
CUDA NEWS
Countdown to SC10
On Nov. 17 at SC10 in New Orleans, NVIDIA chief scientist Bill Dally will deliver a plenary speech titled "GPU Computing: To ExaScale and Beyond."
  - See: www.nvidia.com/sc10

World's Fastest DX11 GPU
NVIDIA announced a new CUDA-enabled GPU for consumers – GeForce GTX 580, the fastest and quietest GPU in its class. For games that feature tessellation – the key feature of DX11– the 512-core GeForce GTX 580 is up to 160 percent faster than the closest competitive product.
  - See: www.nvidia.com

#1 Molecular Graphics Paper
The top most-downloaded paper this week in the Journal of Molecular Graphics and Modeling is "GPU-Accelerated Molecular Modeling Coming of Age" by John Stone, David Hardy, Ivan Ufimtsev and Klaus Schulten (GTC 2010 keynote speaker). Research highlights:
  • GPUs have become powerful accelerators for molecular modeling applications
  • GPUs provide better price-performance than traditional computing techniques
  • GPU clusters consume less space, power and cooling than traditional clusters
  - See: http://is.gd/gLsPl

IMPETUS Afea's Finite Element Code for GPUs
IMPETUS Afea of Norway announced Afea Solver, a non-linear explicit finite element tool. The new GPU-accelerated code can predict deformations of structures exposed to extreme loading conditions. Finite element analysis is widely used in the aeronautical, biomechanical and automotive industries.
  - See: www.impetus-afea.com/?p=6

SciComp's GPU-Accelerated Pricing Software
SciComp enhanced its derivatives pricing software. "The mathematical problems of pricing derivatives are tailor-made for GPU computing," said Curt Randall of SciComp. "GPUs costs are a small percentage of the cost of a grid solution and offer radical reductions in both footprint and power consumption."
  - See: www.scicomp.com/news/press_releases/pr25.html

Parallel Computing on the Desktop with MATLAB
MathWorks held a webinar this week on how to use the Parallel Computing Toolbox to speed up MATLAB apps on GPU-equipped hardware. The webinar was given by Eric Johnson of MathWorks.
  - See: http://is.gd/gQgr0
CUDA CALENDAR
November 2010

– Supercomputing 2010 (SC10)
Nov. 13-19, New Orleans
The NVIDIA GPU Computing Theater at SC10 will feature talks by industry luminaries, scientists and developers. All conference attendees are invited to participate.
www.nvidia.com/sc10

– GPU Programming with CUDA Fortran, CUDA C, PGI Accelerator - PGI (at SC10)
Nov. 15, New Orleans (by Michael Wolfe, PGI)
www.pgroup.com/support/tutorial_outline.htm

– Paving the Road to Exascale - Mellanox (at SC10)
Nov. 17, 7:00 p.m., New Orleans
http://www.mellanox.com/sc10_event/index.php

– Advanced GPU Supercomputing for High-Frequency Trading
Nov. 15-17, New York (by Andrew Sheppard)
http://ajtsheppard.wordpress.com

– Accelerating Matlab with the GPU - SagivTech & Systematics
Nov. 16 & Nov. 17 (two complimentary half-day workshops), Ramat Gan, Israel
www.sagivtech.com/21262.html

– NEW: Best-in-Class FSI Solutions (webinar) - ACUSIM
Nov. 18, 5:30 a.m. pacific
http://is.gd/gLglY

– Training from CAPS
Nov. 23-25, Rennes, France
www.caps-entreprise.com

– Improve Time to Debug (webinar) - Allinea Software
Nov. 24-25. Contact: sales@alinea.com

– MATLAB Expo
Nov. 26, Tokyo
http://matlabexpo.com

– Call for Papers - IEEE/ACM Intl. Symposium on Cluster, Cloud and Grid Computing
Papers deadline: Nov. 30
http://www.ics.uci.edu/~ccgrid11/

December 2010

– CUDA and Advanced Image Processing - SagivTech
Dec. 12-14, Ramat Gan, Israel
www.sagivtech.com/24054.html

– UK GPU Computing Conference - Univ. of Cambridge
Dec. 13-14, Cambridge, UK
http://www.many-core.group.cam.ac.uk/ukgpucc2/

– SIGGRAPH Asia
Dec. 16-18, Seoul
www.siggraph.org/asia2010

2011

– Scientific Computing in the Americas: The Challenge of Massive Parallelism
Jan. 3-14, 2011, Valparaiso, Chile
www.bu.edu/pasi

– IEEE International Parallel & Distributed Processing Symposium
May 16-20, 2011, Anchorage
www.ipdps.org

– NEW: Intelligent Vehicles Conference - IEEE
June 5-9, 2011, Baden-Baden, Germany
http://www.mrt.uni-karlsruhe.de/iv2011/

– NEW: Internat'l. Conference on Computer Systems and Applications
June 27-30, 2011, Sharm El-Sheikh, Egypt
http://aiccsa2011.hpcl.gwu.edu/

Ongoing
– CUDA Certification: www.nvidia.com/certification
– GPU Computing Webinars: www.nvidia.com/webinars
– Training from EMPhotonics: www.emphotonics.com/services/cuda-training

(To list an event, email: cuda_week_in_review@nvidia.com)
CUDA RESOURCES
GPU Technology Conference
– See presentations and keynotes from GTC 2010: www.nvidia.com/gtc
CUDA GPUs
– See list of CUDA-enabled GPUs: www.nvidia.com/object/cuda_gpus.html
CUDA and Parallel Nsight Overview
– See blog post and video: http://is.gd/gbGen
CUDA Downloads
– Download CUDA Toolkit 3.2: http://bit.ly/aKCENp
– Download OpenCL v1.1 pre-release drivers and SDK code samples (Log in or
   apply for an account)
– Get developer guides and docs: http://developer.nvidia.com/object/gpucomputing.html
CUDA and Academia
– Learn more at http://research.nvidia.com/
CUDA on the Web
– See previous issues of CUDA: Week in Review: http://is.gd/cBXbg
– Follow CUDA & GPU Computing on Twitter: www.twitter.com/gpucomputing
– Network with other developers: www.gpucomputing.net
– Stayed tuned to GPGPU news and events: www.gpgpu.org
– Learn more about CUDA on CUDA Zone: www.nvidia.com/cuda
CUDA Recommended Reading
– Read Kudos for CUDA: www.hpcwire.com/features/Kudos-for-CUDA-97889444.html
– Read Supercomputing for the Masses, Part 20: http://is.gd/f9o6o
– Read CUDA books: http://www.nvidia.com/object/cuda_books.html
About CUDA
CUDA is NVIDIA’s parallel computing hardware architecture. NVIDIA provides a complete toolkit for programming on the CUDA architecture, supporting standard computing languages such as C, C++ and Fortran as well as APIs such as OpenCL and DirectCompute. Send comments and suggestions to: cuda_week_in_review@nvidia.com
Stay in Touch with NVIDIA
Twitter   Follow GPU Computing on Twitter
Facebook   Become a fan of NVIDIA on Facebook
NVIDIA online profiles   See list of NVIDIA online profiles

Click here to opt in specifically to CUDA: Week in Review.

Copyright © 2010 NVIDIA Corporation. All rights reserved. 2701 San Tomas Expressway, Santa Clara, CA 95050.