NVIDIA CUDA: Week in Review -- Issue 49
NVIDIA CUDA & GPU computing news from around the world
CUDA SPOTLIGHT
In this week's spotlight, we talked to Sanford Russell of NVIDIA about CUDA 4.0:
NVIDIA: Why is CUDA 4.0 important?
Sanford: When we first released CUDA 1.0 in 2007, it was targeted to early adopters. Since then, it has evolved into a robust software package used around the world. The latest version, CUDA 4.0, is perfect for C/C++ and Fortran parallel developers. We see it being attractive to a broad base of developers, including those who may have been on the fence about parallel computing. Every year, the return on investment for CUDA-based application development becomes more compelling.
NVIDIA: What's new in CUDA 4.0?
Sanford: We have three critical new features, all of which make parallel programming easier and accessible to more developers:
- NVIDIA GPUDirect 2.0, which enables peer-to-peer memory access and thus faster multi-GPU programming;
- Unified Virtual Addressing, which provides a single flat memory address space for CPU and GPU resources, enabling quicker and easier parallel programming;
- More C++ capabilities and an easier way to program parallel apps in C++ with the addition of the Thrust library of template performance primitives.
NVIDIA: When will the software be available?
Sanford: The CUDA Toolkit 4.0 Release Candidate will be available for download on Friday for CUDA Registered Developers. You can register at: www.nvidia.com/paralleldeveloper. The software will be available at: https://developer.nvidia.com/object/cuda_4_0_RC_downloads.html.
- For more info, see Sanford's post on the NVIDIA blog: https://blogs.nvidia.com/
- Register for the CUDA 4.0 webinar (Fri., March 4, by Will Ramey of NVIDIA): https://bit.ly/fiMSzX
CUDA DEVELOPER NEWS
- GPU Computing Meetups Spring Up
- A number of GPU computing "Meetups" have sprung up. These professional groups - which leverage the social media platform Meetup - have formed in New York City, Silicon Valley, Boston, Chicago, Albuquerque and Tokyo in the first eight weeks of 2011. The networks provide a way for developers and business people to get together informally to discuss applications, markets and technologies related to GPU computing. The first GPU Meetups were founded in Australia - in Brisbane, Sydney, Perth and Melbourne.
- - To see a list of GPU-related Meetups, visit: https://gpu.meetup.com/
- Gordon Bell Prize - Application Deadline
- The ACM Gordon Bell Prizes are awarded each year to recognize outstanding achievement in high-performance computing. The purpose of the award is to track the progress over time of parallel computing. Abstracts are due April 1 and papers are due April 8. See: https://awards.acm.org/bell/
- GTC 2011 - Save the Date!
- The next GPU Technology Conference (GTC 2011) will take place October 11-14 in San Jose, California. Call for submissions will open soon. See: https://www.gputechconf.com
Related Articles:
|
