Thursday, April 9th 2009

NVIDIA Introduces Industry's First Debugger and Profiler For GPU Computing

The CUDA architecture continues to blaze a trail as the leading platform for developing and running GPU Computing applications, with support for C, OpenCL, DirectX Compute, Fortran and other languages and APIs. The latest CUDA 2.2 Beta contains a host of significant new features, including:

Hardware debugger for the GPU
Linux developers can now use a debugger on CUDA-enabled GPUs that offers both the familiar interface of the popular open-source GDB debugger and the ability to debug kernels as they execute on the GPU. The GPU-side debugger has all the features that developers expect from GDB, including the ability to have breakpoints, watch variables, inspect state, etc., as well as additional functions for CUDA-specific features.

Visual Profiler v2.2 for the GPU
The most common step in tuning application performance is profiling the application and then modifying the code. The CUDA Visual Profiler is a graphical tool that enables profiling C applications running on the GPU. This latest release of the CUDA Visual Profiler supports full measurement of memory bandwidth within a kernel, giving developers visibility into one of the most performance-critical areas of CUDA.

Full support for Microsoft Windows Server 2003/2008
Tesla C1060 and S1070 are now fully supported under both Microsoft Windows Server 2003 and 2008, offering developers and high-performance computing users more flexibility in their choice of operating system. CUDA 2.2 runs on Windows, MacOSX and major LINUX distributions.

Additional Features Coming with CUDA 2.2
  • Improved OpenGL interop performance
  • Texture from pitch linear memory
  • Zero-copy support for direct access to system memory
  • Pinned shared system memory allows compute kernels to share system memory
  • Asynchronous memcopy on Vista
For a full list of new features in CUDA 2.2 and to get early access, see discussion on the CUDA forums. Source: NVIDIA
Add your own comment

8 Comments on NVIDIA Introduces Industry's First Debugger and Profiler For GPU Computing

wow. nvidia really is serious about cuda.
Posted on Reply
They have to be if they want to beat AMD and Intel when it comes to ATi Stream and Havok
Posted on Reply
The race for CUDA (for graphics and for math) is just like the format wars of HDDVD and BlueRay, or Betamax and VHS, or ATRAC and MP3. It is who gets there first and is adopted first.

To win the CUDA race, nV *must* get their hardware out there NOW and programmers using their tools, so as to win the GPU-coding library wars.

One way to do that is to offer easy to use free developer software and good documentation. Another idea is to HALVE PRICES on enthusiasts boards :D, with special discounts to TPU users :D LOL

They need to do a competitor trade-up program. Hand in your X1, x2 or X3 generation ATI card for a half price nV GTX260 series. That would help their domination.
Posted on Reply
I think Nvidia might be screwed with CUDA, ATI's STREAM has made alot of progress really fast, as soon as ATI gets some developers to utilize it nvidia will really have to get off their toes and I think thats starting to happen now
Posted on Reply
It might be too late, but nV need to get nV CUDA products into the universities and software houses that write and use library code. They also need to make a CUDA warehouse website to open source a lot more stuff. They are doing something, but not enough. If a developer wants to "try out" some CUDA for proof of concept, or to speed up certain code, or whatever, the support, libraries and cookbooks need to be there.

nV should do a "Developer month". Register as a developer, download the tools, access the forum, and should get a CUDA-COUPON for big discounts on the hardware.

And the developer forum should be actively supported with seasoned developers helping with code snippets for n00bies.
Posted on Reply
I admire their efforts. It will be interesting to see if it pays off.
Posted on Reply
little note havok is OpenCL based for gpu acceleration, and as i understand it, current nvidia beta's support opencl.....hence we SHOULD beable to accelerate HavoK with any cuda compatable card.
Posted on Reply