Site Archive (Complete)
DrDobbs Portal Blog: CUDA, WUDA, SHUDA
EDITOR'S EYE

The World of Software Development.

by Jon Erickson
November 14, 2007

CUDA, WUDA, SHUDA

Even if you're a hardcore C programmer, there's every chance that one C compiler you haven't heard about is CUDA. That is, unless you're a game programmer or maybe a college student, not that the two have anything in common. But more on that later.

Actually, CUDA is more than just a C compiler. It is an SDK that includes a C compiler, hardware debugger, and a performance profiler. And both CUDA 1.1 for Windows, which supports 64-bit Windows XP, and CUDA 1.1 for Linux are freely available for downloading. CUDA's new debugger is based on a standard GDB interface and lets you set breakpoints, set/examine variables directly on the processor, supports a standard DDD graphical interface, and allows third-party debuggers to seamlessly layer on top of CUDA. The SDK also includes standard FFT and BLAS libraries.

The reason you probably haven't heard of CUDA is that it comes from an unlikely source. Unlikely, that is, in terms of the usual suspects when you think of software vendors. CUDA is produced by NVIDIA in support of its graphics processing unit (GPU). Not that everyone developing for NVIDIA's GPU is creating games. You can bet, however, that they're building some kind of data-centric application, of which graphics are the most demanding.

So why would students be more likely to know about CUDA than your average C programmer? Because a growing number of schools including Stanford, UC Berkeley, Johns Hopkins, and the India Institute of Technology, among others are using CUDA as core curriculum in their parallel-programming courses.

It's like John Owens, an assistant professor at the University of California, Davis said: "Perhaps the most important challenge facing the computing community is the move to parallel processing. As educators, teaching parallel hardware and software today are vital to giving our students the tools they need to build tomorrow's hardware and software. NVIDIA GPUs and the CUDA programming environment are a terrific way for us to put cheap, powerful data-parallel processing on the desktop for all our students."

That's okay. And the price is right, too.

-- Jonathan Erickson
jerickson@ddj.com

Posted by Jon Erickson at 06:18 PM  Permalink





January 2008
Sun Mon Tue Wed Thu Fri Sat
    1 2 3 4 5
6 7 8 9 10 11 12
13 14 15 16 17 18 19
20 21 22 23 24 25 26
27 28 29 30 31    


BLOGROLL
 
INFO-LINK


Related Sites: DotNetJunkies, SD Expo, SqlJunkies