Announcement

Collapse
No announcement yet.

A Big Comparison Of The AMD Catalyst, Mesa & Gallium3D Drive

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Wyatt
    replied
    Originally posted by marek View Post
    No, I think this is the only one.
    What about S3TC? I seem to recall it had similar problems (which is annoying as all crap).

    Leave a comment:


  • marek
    replied
    Originally posted by Wyatt View Post
    How many of these "patented but implemented" feature branches are floating around anyway? Has anyone taken the time to get up a git repo that tracks and merges them as well as possible?
    No, I think this is the only one.

    Leave a comment:


  • Wyatt
    replied
    Originally posted by marek View Post
    The Mesa maintainers appear not to want patented features in master and they ignore any of my patches that might be infringing in US. The floating-point formats from GL_ARB_texture_rg can only be used if GL_ARB_texture_float is available as well, the support of which is here:
    http://cgit.freedesktop.org/~mareko/mesa/?h=floating2
    How many of these "patented but implemented" feature branches are floating around anyway? Has anyone taken the time to get up a git repo that tracks and merges them as well as possible?

    Leave a comment:


  • marek
    replied
    Originally posted by oibaf View Post
    would be possible to integrate the new cs code into libdrm so that also the ddx (and eventually also r600g, and classic radeon, r200, r300 and r600) can benefit from it?
    I don't think so. My changes are rather radical and I don't want to destabilize libdrm. Also, r600g doesn't use libdrm.

    Originally posted by oibaf View Post
    The Mesa maintainers appear not to want patented features in master and they ignore any of my patches that might be infringing in US. The floating-point formats from GL_ARB_texture_rg can only be used if GL_ARB_texture_float is available as well, the support of which is here:

    Leave a comment:


  • Azpegath
    replied
    Originally posted by BlackStar View Post
    1. VSynced rendering (nvidia tears without compiz).
    2. Desktop zoom when trying to read text on your HDTV across your living room.
    3. Shadow decorations to help pick windows apart.
    4. Better performance on fglrx than regular 2d rendering.
    5. The window scale plugin which is like alt-tab on serious overdrive.
    I'm glad that it works so well for you, but to be honest it hasn't really ever been stable for me. Neither with the Catalyst nor the open drivers.

    1. Why worry about Nvidia when I'm using ATI? =)
    2. Hm, not really a use-case for me, but I guess you use it
    3. I feel like I'm perfectly capable of telling windows apart without shadow decorations
    4. This is a very good reason. This is actually the reason why I was so excited about "GPU accelerated desktops" in the first place. But now I feel that they should be renamed to "GPU decelerated desktops".
    5. Sure, it's bling bling, but is it really the most exciting feature? Can't we do more with the technology than this?

    I just feel like the "killer feature" hasn't arrived yet, and sadly I don't have a good idea of what it could be either. If I did, I'd try to implement it, or at least communicate it to the community and discuss it.
    I think a true acceleration is awesome, but it just doesn't seem to get a whole lot better when enabling it. The only difference I see it that the CPU load is severely decreased when frantically moving a window back and forth.

    Leave a comment:


  • oibaf
    replied
    Marek,

    Originally posted by marek View Post
    r300g has a faster winsys since yesterday (commit).
    would be possible to integrate the new cs code into libdrm so that also the ddx (and eventually also r600g, and classic radeon, r200, r300 and r600) can benefit from it?

    Also is there a plan to merge the floating-point formats from GL_ARB_texture_rg patches?

    Leave a comment:


  • agd5f
    replied
    you can try 'git fetch --all', but either should work.

    Leave a comment:


  • RealNC
    replied
    Originally posted by agd5f View Post
    Here's the basics (not sure this is 100% right, doing this from memory):
    #clone Linus tree:
    git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
    cd linux-2.6
    # check out 2.6.37 on a new branch called my-2.6.37
    git checkout -b my-2.6.37 v2.6.37
    #add Dave's tree:
    git remote add airlied git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git
    # fetch the updates from Dave's tree
    git fetch airlied
    # merge drm-next into the current branch
    git pull airlied/drm-next
    The above just hangs at "git fetch airlied". No CPU load and no network activity.

    Leave a comment:


  • BlackStar
    replied
    Originally posted by Azpegath View Post
    +1

    I'm not using Compiz for two simple reasons:
    1) It's to big a hit on performance
    2) I still haven't seen a good (enough) use for it
    1. VSynced rendering (nvidia tears without compiz).
    2. Desktop zoom when trying to read text on your HDTV across your living room.
    3. Shadow decorations to help pick windows apart.
    4. Better performance on fglrx than regular 2d rendering.
    5. The window scale plugin which is like alt-tab on serious overdrive.

    Leave a comment:


  • Azpegath
    replied
    Originally posted by Azpegath View Post
    Perhaps a stupid question, but why aren't the FOSS drivers multi-threaded yet? It feels like a shame to me, sitting on an Core i7.
    Nevermind, I've now gone through the 19 pages of comments and got my answer very well answered by bridgman

    Leave a comment:

Working...
X