[AMD] Close to the Metal: Any newer Guides than 1.01 - possibly with R600-stuff in it

Discussion in 'GPGPU Technology & Programming' started by CarstenS, Sep 24, 2007.

  1. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,798
    Likes Received:
    2,056
    Location:
    Germany
    Is there such a thing already? I seem to be able to find only 2006's version 1.01.


    Thanks for any heads-up!
     
  2. Tim Murray

    Tim Murray the Windom Earle of mobile SOCs
    Veteran

    Joined:
    May 25, 2003
    Messages:
    3,278
    Likes Received:
    66
    Location:
    Mountain View, CA
    Nothing that has been released publicly as of yet.
     
  3. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,798
    Likes Received:
    2,056
    Location:
    Germany
    Damn, but thanks anyway :)
     
  4. Anarchist4000

    Veteran Regular

    Joined:
    May 8, 2004
    Messages:
    1,439
    Likes Received:
    359
    Any idea what happened to the CTM stuff? Seems like it sort of died off. I think their site moved or something but I never heard anything about it.
     
  5. silent_guy

    Veteran Subscriber

    Joined:
    Mar 7, 2006
    Messages:
    3,754
    Likes Received:
    1,379
    I don't think it matters: how many are really willing to dive into this kind of low-level hackery to extract the ultimate amount of performance, when you can get similar or close enough performance writing shaders or C? I stopped writing ASM somewhere in the mid nineties and don't really feel like going back. :wink:
     
  6. Tim Murray

    Tim Murray the Windom Earle of mobile SOCs
    Veteran

    Joined:
    May 25, 2003
    Messages:
    3,278
    Likes Received:
    66
    Location:
    Mountain View, CA
    Well, you could compile HLSL to CTM, although if you want any of the memory format stuff you have to fiddle with the ASM output. But yeah, if you're going to GPGPU and your choices are (for the sake of argument) a 120x speedup with R600 and CTM or a 100x speedup with G80, CUDA and much lower development costs, pretty sure most people will go with CUDA. AMD has dropped the ball here.
     
  7. AlexV

    AlexV Heteroscedasticitate
    Moderator Veteran

    Joined:
    Mar 15, 2005
    Messages:
    2,528
    Likes Received:
    107
    The GPGPU dudes care about it, and I think both this and CUDA are aimed at them in particular. For all of us other heathens, I agree with you though.
     
  8. MfA

    MfA
    Legend

    Joined:
    Feb 6, 2002
    Messages:
    6,806
    Likes Received:
    473
    Enough people wanted access to the lower levels for NVIDIA to expose them (PTX). They both dropped the ball, NVIDIA Just picked it up. AFAICS AMD/ATI is intending on providing a higher level interface in the future (of course there is already commercially available development software which does that if you don't want to touch CTM).
     
    #8 MfA, Sep 26, 2007
    Last edited by a moderator: Sep 26, 2007
  9. mhouston

    mhouston A little of this and that
    Regular

    Joined:
    Oct 7, 2005
    Messages:
    344
    Likes Received:
    38
    Location:
    Cupertino
    Response about CTM status from AMD:

    http://www.gpgpu.org/forums/viewtopic.php?t=4789

    In short, the software stack is being redone as talked about in the GPGPU course at SIGGRAPH and a better separation between the hardware layer and the compute layer is being done.
     
  10. silent_guy

    Veteran Subscriber

    Joined:
    Mar 7, 2006
    Messages:
    3,754
    Likes Received:
    1,379
    From reading the CUDA forums, I can see that some people indeed want to go below the C level, though mostly to understand what's going on rather than get their hands dirty, which indeed seems sensible to me.

    Moving CTM up the abstraction chain is nice, but as long as it's not C-like, it will be incredibly difficult to compete. I don't play games and the only reason I bought a GF8 card is to use CUDA. I assume a lot of others (universities, companies) will think the same way.

    If ATI wants to regain GPGPU mindshare they'll have to come up with a decent C equivalent. Maybe that's what they are about to announce soon?
     
  11. Tim Murray

    Tim Murray the Windom Earle of mobile SOCs
    Veteran

    Joined:
    May 25, 2003
    Messages:
    3,278
    Likes Received:
    66
    Location:
    Mountain View, CA
    I assume this is the Compute Abstraction Layer that was mentioned some months ago. Not sure if AMD's hired anyone in particular who's built these kinds of things in the past, so I don't know what to expect...
     
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...