Search Unity

  1. Megacity Metro Demo now available. Download now.
    Dismiss Notice
  2. Unity support for visionOS is now available. Learn more in our blog post.
    Dismiss Notice

"SGX: Microkernel unresponsive" and "Split Scene" issue that only occurs on iPad 2

Discussion in 'iOS and tvOS' started by BrianC, Apr 26, 2011.

  1. BrianC

    BrianC

    Joined:
    Mar 3, 2011
    Posts:
    7
    I'm writing an iOS game using OpenGL ES 2 and I encountered an error that only seems to occur on the iPad 2. The message "SGX: Microkernel unresponsive" is printed to the console and there is a delay of 1010 milliseconds (on average) until the next frame. The OpenGL ES Driver Instrument displays a "Split Scene Count" of 5 when the error messages is printed. The error occurs in a variety of situations and doesn't always occur at the same time. It occurs more frequently when 4x MSAA and alpha blending are enabled, but it also occurs when they are disabled.

    I've run my test program on two iPad 2 devices with iOS 4.3 and 4.3.2 and the same error occurred on both. The error did not occur on an original iPad (iOS 4.3), an iPod Touch 4 (iOS 4.2.1), and an iPhone 3GS (iOS 4.2.1).

    I'm having trouble finding more information about both the "SGX: Microkernel unresponsive." error and the split scene message. Does anyone know what they mean?

    From the iPad 2 console:

    Code (csharp):
    1.  
    From Instruments:
    $InstrumentsScreenshot1.png
     
  2. Alexey

    Alexey

    Unity Technologies

    Joined:
    May 10, 2010
    Posts:
    1,623
    Seems like we found some possible causes
    Can you please bug report with repro project so we can test our fix, so it can go live with 3.4?
    Drop the case number here (6 digits number, no links)
     
  3. BrianC

    BrianC

    Joined:
    Mar 3, 2011
    Posts:
    7
    Thanks Alexey. I'm working to create a simpler project that has the bug so that I can report it. What do you think the possible causes are? That information could help me make a test project more easily.
     
  4. Alexey

    Alexey

    Unity Technologies

    Joined:
    May 10, 2010
    Posts:
    1,623
    Well, i can't say for sure, just guesses (and all of them internal to unity), but the fact that you did poked guys in apple forums is good - sometime they are faster then us in tracking down some crazy driver problems ;-))
    P.S. just to make sure - did you try to disable plugins if you have any for repro? we saw some crazy bugs when using crazy plugins ;-)
     
  5. BrianC

    BrianC

    Joined:
    Mar 3, 2011
    Posts:
    7
    I'm still working to create the sample project that reproduces the issue. I'm not using any plugins. Everything is rendered using a shader that outputs a constant color.

    Do you know what a BIF fault is? The console messages in my initial post state that a BIF fault was found. What could be causing this?

    Apr 26 01:06:39 unknown kernel[0] <Debug>: Found BIF fault
    Apr 26 01:06:39 unknown kernel[0] <Debug>: device address: 0x19F4C000 (gart offset: 415 MB)
    Apr 26 01:06:39 unknown kernel[0] <Debug>: requestor: DCU
    Apr 26 01:06:39 unknown kernel[0] <Debug>: Page directory entry: 0x8C01A001********* Page table entry: 0x00000000

    It might be related to this post on the Apple forums.

    I really hope we can figure this out so that it can be fixed in 3.4.
     
  6. Alexey

    Alexey

    Unity Technologies

    Joined:
    May 10, 2010
    Posts:
    1,623
    well, gart = Graphics Address Remapping Table, meaning that "gart offset: 415 MB" is most likely too much
    usually these kind of errors (well, judging from discussions found with google) are due to some code telling gl to read out of bounds of some buffer.
    So, btw, if you are using smth from GL class - try to remove it too
     
  7. BrianC

    BrianC

    Joined:
    Mar 3, 2011
    Posts:
    7
    Thanks for your quick reply. I'm removing some code that used the GL class and testing it again.
     
  8. Wozik

    Wozik

    Joined:
    Apr 10, 2009
    Posts:
    662
    BrianC, you may still want to report a bug to check out the fix of a possible bug ;-)
     
  9. BrianC

    BrianC

    Joined:
    Mar 3, 2011
    Posts:
    7
    Case number 400892
     
  10. Wozik

    Wozik

    Joined:
    Apr 10, 2009
    Posts:
    662
    BrianC, great thanks. The bug has a developer working on a fix.
     
  11. cactusov

    cactusov

    Joined:
    Oct 12, 2011
    Posts:
    1
    Hi Wozik,
    Could I know what about this bug, is it fixed?