CUDA: Week In Review

CUDA Pro Tip: Vectorized memory accesses are a fundamental CUDA optimization that can be used to increase bandwidth utilization while reducing instruction count latency. Learn more atParallel Forall.

GPU Tech Conference Update: GTC 2014 registration is now open. Special code for newsletter readers: GM20CD. Call for Posters is open through Jan. 31.

CUDA SPOTLIGHT
CUDA-Accelerated Cosmology This week’s Spotlight is on Dr. Debbie Bard of the Kavli Institute for Particle Astrophysics and Cosmology. KIPAC is an independent laboratory of Stanford University. Its mission is to bring the resources of modern computational, experimental, observational and theoretical science to bear on our understanding of the universe at large. Read the Spotlight.
CUDA NEWS
SC13 Wrap Up SC13 in Denver attracted 10,600+ attendees and 335 exhibitors. Highlights included the Student Cluster competition, won by a team from the University of Texas at Austin using NVIDIA GPUs. NVIDIA’s SC13 news announcements included: Tesla K40: World’s Fastest Accelerator for Supercomputing CUDA 6 Dramatically Simplifies Parallel Programming IBM and NVIDIA Partner to Build Next-Gen Supercomputers Student-Led CUDA Lab at Virginia Tech Kudos to the ACM Student Chapter at Virginia Tech for hosting a CUDA tutorial and lab. OrganizerCarlo del Mundo led students through hands-on exercises using the cloud-based qwikLABS platform, which makes several thousand GPU-powered Amazon EC2 servers available for online training. Students learned to write and launch CUDA C/C++ kernels, manage GPU memory, and manage communication and synchronization. If you are a student or teacher and would like to offer a GPU programming lab at your university, contact cuda-cloud@nvidia.com. Compute the Cure Grant Winner More than 50 universities and research organizations from around the world recently submitted proposals for a $200,000 grant from the NVIDIA Foundation. The winner is Dr. Rommie Amaro, University of California, San Diego. Dr. Amaro is working on a shareable GPU-accelerated workflow to help speed the development of drugs to fight cancer. This grant is part of the Foundation’s Compute the Cure initiative. CUDA + Adobe The newest release of Adobe Creative Cloud adds even more GPU-accelerated solutions for creative video pros including debayering in Premiere Pro CC for ultra-high definition 4K formats, a new Direct Link Color Pipeline in SpeedGrade CC, and GPU-enhanced Adobe Media Encoder for faster video and image processing. Test Drive Tesla K40 Today With cutting-edge features like GPUBoost and 12GB of memory, Tesla K40 delivers up to 40% more performance than Tesla K20X. You can try Tesla K40 for free by registering for a GPU Test Drive. Upcoming Webinars Dec. 12: Running OpenACC Programs, M. Wolfe, PGI Dec. 12: GPU-Accelerated Geospatial Line-of-Sight Calculations, F. Suykens, B. Adams, Luciad Dec. 17: NVIDIA GRID VCA: High-Performance Remoting Solution for SolidWorks, A. Patel, NVIDIA Jan. 16: Folding@home and OpenMM: Simulating Protein Dynamics, V. Pande, Stanford University Jan. 21: Using HOOMD-blue for Polymer Simulations and Big Systems, J.Glaser, Univ. of Michigan
NEW ON THE BLOG	back to the top
Parallel Forall: CUDACasts Episode #12: Programming GPUs using CUDA Python, by Mark Ebersole CUDA Pro Tip: Increase Performance with Vectorized Memory Access, by Justin Luitjens
NVIDIA: 12-Year-Old Learns to Program Graphics Chips, by Brian Caulfield GPU-Powered System Wins Student Supercomputing Competition, by Mark Ebersole More Than Skin Deep: P&G Using GPUs to Study Skincare Products, by George Millington How One Researcher Used 18,000 GPUs..., by Brian Caulfield
CUDA JOB OF THE WEEK	back to the top
NVIDIA seeks senior software engineer to work on CUDA system software team. Ideal candidate will have very strong C programming skills, thorough understanding of operating systems and kernel programming, working knowledge of CUDA and parallel programming, a good understanding of hardware architecture, and excellent communication and planning skills. Seewww.nvidia.com/careers, #1635268.
CUDA EDUCATION	back to the top
Want to improve your technical skills? Sign up for Intro to Parallel Programming. Need CUDA advice? See list of worldwide CUDA trainers and consultants. Have CUDA questions? Check out NVIDIA DevTalk forums and Stack Overflow. Require fast access to docs? Visit the CUDA doc library.
CUDA CALENDAR	back to the top
December 4-Day CUDA Course – Finance Focus (Acceleware)   Dec. 10-13, 2013, New York, New York UK Many-Core Developer Conference (University of Oxford)   Dec. 16, 2013, Oxford, England Note: Includes CUDA "masterclass" sessions Parallel and Distributed Computing in Geoscience and Remote Sensing   Dec. 15-18, 2013, Seoul, Korea Note: IEEE Workshop 2014 CUDA Programming Course (Delft University of Technology)   Jan. 24, 2014, Delft, Netherlands Teachers: Prof.dr.ir. C. Vuik and Ir. C.W.J. Lemmens ASPLOS 2104   March 1-5, 2014, Salt Lake City, Utah GPU Technology Conference (GTC 2014)   March 24-27, 2014, San Jose, Calif. IEEE Int’l Parallel & Distributed Processing Symposium   May 19-23, 2014, Phoenix, Arizona (To list an event, email: cuda_week_in_review@nvidia.com)
CUDA RESOURCES	back to the top
Online CUDA Course in Russian Try out this new course designed for Russian speakers. CUDACasts Check out our new series of short videos about CUDA. Tell Us Your CUDA Story If you are a CUDA developer, tell us how you are using CUDA. Your submission will be considered for display on the NVIDIA website. GPU-Accelerated Apps See updated list of 240+ GPU-accelerated applications. GPU Meetups Learn about Meetups in your city, or start one up. The Best of GTC 2013, Online GTC 2013 featured over 400 sessions on breakthroughs made with GPUs in science, technology, and industry. Experience it by visiting http://www.nvidia.com/gtc2013. GPU Test Drive Want to try Tesla K20 for free? Sign up here. CUDA Documentation The new CUDA documentation site includes release notes, programming guides, manuals and code samples. Online Learning Udacity | Coursera NVIDIA Developer Forums Join us on the NVIDIA DevTalk forums to share your experience and learn from other developers. You can also ask questions on Stack Overflow, using the ’cuda’ tag. CUDA Consulting Training, programming and project development services are available from CUDA consultants around the world. To be considered for inclusion on list, email: cuda_week_in_review@nvidia.com. GPU Computing on Twitter For daily updates about GPU computing and parallel programming, follow @gpucomputing on Twitter. Downloads CUDA 5 Nsight CUDA on the Web CUDA Spotlights CUDA Newsletters CUDA Zone GPU Test Drive GPUComputing.net NVIDIA Newsletters Sign up for NVIDIA Newsletters, including Media & Entertainment, GeForce, NVIDIA GRID VDI and Shield.