Page 1 of 2 12 LastLast
Results 1 to 10 of 12

Thread: Linux 2.6.33-rc7 Released, Still Many Regressions

  1. #1
    Join Date
    Jan 2007
    Posts
    14,373

    Default Linux 2.6.33-rc7 Released, Still Many Regressions

    Phoronix: Linux 2.6.33-rc7 Released, Still Many Regressions

    Linus pushed out the Linux 2.6.33-rc7 kernel yesterday afternoon, but still even being seven release candidates into the Linux 2.6.33 kernel cycle, at least one more test release is due before we may see an official release. The number of regressions for the Linux 2.6.33 kernel is still high and is somthing that isn't pleasing Torvalds. While the Linux 2.6.33-rc7 kernel release should lower the regression count with many fixes going in since 2.6.33-rc6 last Friday, there is also some Intel and ATI Radeon DRM driver updates still being pushed into this next major release...

    http://www.phoronix.com/vr.php?view=Nzk2Ng

  2. #2
    Join Date
    Oct 2007
    Posts
    50

    Default

    Michael I think the kernel-tracker.phoromatic.com is a verry good ideea, hope the kernel developers take a look from time to time at it.

  3. #3
    Join Date
    Jan 2009
    Location
    Columbus, OH, USA
    Posts
    323

    Default

    Speaking of regressions on kernel-tracker, has the Threaded I/O Tester really gotten that much faster as of a week ago? A sudden two-order speedup seems slightly suspicious to me. Is it actually doing the test?

    As a side note, it seems it would be handy to have a "Most Recent: " result on the PTS graphs in addition to average, peak, and low. Or even just number along the left side to match the right.

  4. #4

    Default

    Quote Originally Posted by Wyatt View Post
    Speaking of regressions on kernel-tracker, has the Threaded I/O Tester really gotten that much faster as of a week ago? A sudden two-order speedup seems slightly suspicious to me. Is it actually doing the test?
    No, it's a new version of the test profile and due to a bug it's not showing up under a new graph, but will be fixed when I get back to my office.

    Quote Originally Posted by Wyatt View Post
    As a side note, it seems it would be handy to have a "Most Recent: " result on the PTS graphs in addition to average, peak, and low. Or even just number along the left side to match the right.
    Good point, will add that when back to office.

  5. #5
    Join Date
    Apr 2009
    Posts
    519

    Default

    I guess this post is too late to the game to be seen by anybody. But starting with rc7 I lost 3D acceleration in my ATI HD3450. It worked pretty well up until rc6 (this is all kubuntu lucid). And it worked 25% faster than the Ubuntu 2.6.32 kernel.

  6. #6
    Join Date
    Oct 2007
    Location
    Toronto-ish
    Posts
    7,386

    Default

    Can you go back and forth between rc6 and rc7 with/without acceleration (to make sure it wasn't some other component making the difference) ? If so suggest you file a bug ASAP, or check on #dri-devel to see how the devs want to handle it.

    Any new messages in dmesg output or xorg log ?

    What exacttly does "lost acceleration" mean in this case ?

  7. #7
    Join Date
    Apr 2009
    Posts
    519

    Default

    Quote Originally Posted by bridgman View Post
    Can you go back and forth between rc6 and rc7 with/without acceleration (to make sure it wasn't some other component making the difference) ? If so suggest you file a bug ASAP, or check on #dri-devel to see how the devs want to handle it.

    Any new messages in dmesg output or xorg log ?

    What exacttly does "lost acceleration" mean in this case ?

    yes, I can go back and forth. This is with the open source stack. I haven't had the time to check dmesg or the xorg log systematically as to file a decent bug report.

    What I mean is that if I fire up any 3D game or PTS test I get essentially a horribly slow slidewhow at 0 fps (seems to be a frame every 3 to 4 seconds).

    I agree, we want to catch these regressions, I am not sure where to report though. I thought of using the kernel bugzilla, then Drivers / Video (DRI/Non-Intel). Would that make sense? I'll need to get some more info later today in order to do that. I also tried the latest kernel from the mainline ppa (nightly build) and the regression is still there.

    Thanks a lot!

  8. #8
    Join Date
    Oct 2007
    Location
    Toronto-ish
    Posts
    7,386

    Default

    I would go with bugs.freedesktop.org, pick "DRI" then component "DRM/Radeon".

    It's possible the kernel tracker is now the right place but from the older pick lists I'm guessing fd.o is still the way to go.

  9. #9
    Join Date
    Apr 2009
    Posts
    519

    Default

    Quote Originally Posted by bridgman View Post
    I would go with bugs.freedesktop.org, pick "DRI" then component "DRM/Radeon".

    It's possible the kernel tracker is now the right place but from the older pick lists I'm guessing fd.o is still the way to go.
    Thanks, I'll also check rc8, it's just out:
    http://lwn.net/Articles/374196/

  10. #10
    Join Date
    Apr 2009
    Posts
    519

    Default

    I found the issue. Up until rc6 KMS was failing, but starting with rc7 it succeeds. However, with KMS enabled 3D acceleration is incredibly poor. If I disable it from the command line I get performance back.

    But all of this is with Ubuntu Lucid, which is using:

    Code:
    Display Server: X.Org Server 1.7.3.902 (1.7.4 RC 2)
    Display Driver: radeon 6.12.99
    Maybe I need newer radeon for that kernel? If you happen to read this, do you still think it makes sense to file a bug report?

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •