Introducing NVIDIA CUDA

Discussion in 'Beyond3D Articles' started by Arun, Feb 16, 2007.

  1. Arun

    Arun Unknown.
    Moderator Legend Veteran

    Joined:
    Aug 28, 2002
    Messages:
    5,023
    Likes Received:
    299
    Location:
    UK
    <a href="http://www.beyond3d.com/articles/cuda-intro/"><img border="1" src="http://www.beyond3d.com/includes/images.php?id=243" align="right" width="75" height="75"></a>Today, NVIDIA released the beta version of their CUDA SDK, which is their new solution for general-purpose computation on GeForce 8 Series GPUs. It exposes the programmable shading hardware through extensions to ANSI C, with only a small number of deviations from the standard. Overall, it should also be both more efficient and easier to use than previous GPGPU solutions.

    Because this subject might also interest a wider and less technical audience, we've got two versions of the article running today. The first <a href="http://www.beyond3d.com/articles/cuda-intro/">looks at the technology</a> and its implementation relatively in-depth, while the latter is an <a href="http://www.beyond3d.com/articles/cuda-quick/">quick summary</a> of the former. So, no matter which version interests you most, read on!
     
  2. digitalwanderer

    digitalwanderer Dangerously Mirthful
    Legend

    Joined:
    Feb 19, 2002
    Messages:
    17,230
    Likes Received:
    1,752
    Location:
    Winfield, IN USA
    Bless you! :D
     
  3. Otto Dafe

    Regular

    Joined:
    Aug 11, 2005
    Messages:
    400
    Likes Received:
    59
    Hear, hear(or is it here, here?).

    These things have such an insane amount of processing power it seems a shame that it just goes to video games(not that there's anything wrong with video games). I think NV should be lauded for this move, and I might have to cut into my beer money and get one of these suckers to play around with. I really like the fact that they went with a reasonably high level language, I honestly just can't be bothered with asm, unless it's a very specific bottleneck. Besides, the compiler usually knows more about the architecture than I care to.

    Good article, as always.
     
    #3 Otto Dafe, Feb 17, 2007
    Last edited by a moderator: Feb 17, 2007
  4. Anarchist4000

    Veteran Regular

    Joined:
    May 8, 2004
    Messages:
    1,439
    Likes Received:
    359
    Slightly off topic but for CTM you can use HLSL to write the functions and then just copy out the assembly that is generated.

    For instance I've been using the AMD shader analyzer to code some stuff in HLSL, then copy the D3D assembly it spits out and use that for my program. It does take a bit of extra work if you try to do something outside the capabilities of SM3 however but HLSL can be quite useful for relatively simple vector based applications.

    In the future I'm sure there will be some basic libraries released to perform the more basic functions and wrap up CTM.
     
    Jawed likes this.
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...