Announcement

Collapse
No announcement yet.

requested features that got implemented in fglrx

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

  • requested features that got implemented in fglrx

    Hi @all,

    I just wanted to say @bridgman & AMD/ATI:

    you guys are made of pure awesomeness

    because:

    finally I can use a kernel with preemptive rcu, sqlb (slab-allocator) and all the other goodies (BFS, BFQ)


    from my remembrance this wasn't possible a few months ago so this is the proof that AMD continously is improving their driver in subtle ways and also are listening to their customers' needs

    you can't compare their efforts directly with nvidia's because nvidia's linux-driver crew has much more man-power (any confirmation / insight ?)


    So therefore this thread shall be dedicated to the successful efforts of the hardworking linux driver-developers at AMD/ATI

    so now lets start with some features:

    * sqlb (slab-allocator) -> it at least works :P
    * preemptive rcu and/or fully preemptible kernel support -> it at least works and/or sometimes hangs and continously throws some errors:
    [ 698.498926] BUG: using smp_processor_id() in preemptible [00000000] code: X/20241
    [ 698.498944] caller is KCL_MODULE_DecUseCount+0x35/0x50 [fglrx]
    [ 698.498946] Pid: 20241, comm: X Tainted: P 2.6.31-zen7_sqlb #2
    [ 698.498948] Call Trace:
    [ 698.498950] [<ffffffff8136b63c>] debug_smp_processor_id+0xcc/0xe0
    [ 698.498970] [<ffffffffa01421b5>] KCL_MODULE_DecUseCount+0x35/0x50 [fglrx]
    [ 698.498989] [<ffffffffa0142cd1>] KCL_SPINLOCK_STATIC_Grab+0x31/0x100 [fglrx]
    [ 698.499009] [<ffffffffa01433fd>] KAS_InterlockedList_InsertAtTail+0x6d/0x150 [fglrx]
    [ 698.499029] [<ffffffffa014449b>] ? KAS_Event_WaitForEvent+0x26b/0x530 [fglrx]
    [ 698.499054] [<ffffffffa015b4a3>] MCIL_InterlockedList+0x193/0x2a0 [fglrx]
    [ 698.499079] [<ffffffffa015ccbb>] ? MCIL_ZeroMemory+0x4b/0x50 [fglrx]
    [ 698.499113] [<ffffffffa02002c8>] ? FIFO_Queue_InsertAtTail+0x48/0x80 [fglrx]
    [ 698.499128] [<ffffffffa015d147>] MCIL_InterlockedDecrement+0x87/0xa0 [fglrx]
    [ 698.499182] [<ffffffffa020194a>] ? UnRegisterIRQClient_Worker+0xaa/0xe0 [fglrx]
    [ 698.499222] [<ffffffffa0209bb6>] ? CallbackQueueAccess+0x216/0x2a0 [fglrx]
    [ 698.499257] [<ffffffffa02018a0>] ? UnRegisterIRQClient_Worker+0x0/0xe0 [fglrx]
    [ 698.499290] [<ffffffffa015ce42>] ? MCIL_bMiniportCapEnabled+0x82/0xa0 [fglrx]
    [ 698.499331] [<ffffffffa0201223>] ? UnRegisterIRQClient+0x83/0xe0 [fglrx]
    [ 698.499372] [<ffffffffa01fe691>] ? IRQMGR_Access+0x131/0x190 [fglrx]
    [ 698.499410] [<ffffffffa0173e43>] ? fireglAsyncioUnregisterIntMsgHandlers+0x2c3/0x3c0 [fglrx]
    [ 698.499416] [<ffffffff816b03a2>] ? _spin_unlock_irqrestore+0x12/0x40
    [ 698.499460] [<ffffffffa01a0954>] ? asyncIONotifyMsg+0x234/0x3e0 [fglrx]
    [ 698.499492] [<ffffffffa0173079>] ? firegl_asyncio_write+0x189/0x250 [fglrx]
    [ 698.499519] [<ffffffffa0145f9b>] ? ip_firegl_write+0x5b/0xa0 [fglrx]
    [ 698.499526] [<ffffffff810c4f18>] ? vfs_write+0xb8/0x1a0
    [ 698.499533] [<ffffffff8102cd75>] ? do_page_fault+0x125/0x260
    [ 698.499540] [<ffffffff810c580c>] ? sys_write+0x4c/0x80
    [ 698.499549] [<ffffffff8100b5ab>] ? system_call_fastpath+0x16/0x1b
    * VT-switching doesn't lead to hardlock -> in most cases it doesn't hardlock
    * restarting X and or killing it by ctrl+alt+backspace doesn't lead to hardlock -> doesn't for me in most cases
    * add video acceleration support -> partially accomplished (see UVD2)

    now you go on

  • #2
    Originally posted by kernelOfTruth View Post
    Hi @all,

    I just wanted to say @bridgman & AMD/ATI:

    you guys are made of pure awesomeness

    because:

    finally I can use a kernel with preemptive rcu, sqlb (slab-allocator) and all the other goodies (BFS, BFQ)


    from my remembrance this wasn't possible a few months ago so this is the proof that AMD continously is improving their driver in subtle ways and also are listening to their customers' needs

    you can't compare their efforts directly with nvidia's because nvidia's linux-driver crew has much more man-power (any confirmation / insight ?)


    So therefore this thread shall be dedicated to the successful efforts of the hardworking linux driver-developers at AMD/ATI

    so now lets start with some features:

    * sqlb (slab-allocator) -> it at least works :P
    * preemptive rcu and/or fully preemptible kernel support -> it at least works and/or sometimes hangs and continously throws some errors:
    * VT-switching doesn't lead to hardlock -> in most cases it doesn't hardlock
    * restarting X and or killing it by ctrl+alt+backspace doesn't lead to hardlock -> doesn't for me in most cases
    * add video acceleration support -> partially accomplished (see UVD2)

    now you go on
    hi.. on ma testing systems the radeon rock the catalyst out..

    some exampels:

    lastest firefox and lastest DU Kubuntu works on Radeon/R300 and intel IGP but not on Catalyst9.10 8.662 ..............

    Kubuntu 9.1o works much faster on radeon. some programms are not useably on catalyst 9.10 like krita very option in the filters or other stuff hang for 2 seconds (no composit bug .. composit is off)... workflow is bad...

    catalyst9.10 is realy only for gnome and kde3,5 not for kde4+
    Last edited by Qaridarium; 11-05-2009, 10:58 PM.

    Comment


    • #3
      Q, you were supposed to say something nice about fglrx

      Comment


      • #4
        Originally posted by bridgman View Post
        Q, you were supposed to say something nice about fglrx
        O Yes pre-alpfa 10.04 ubuntu use kernel 2.6.32 3D Radeon support...

        in the past i chance my systems to 9.10 direct after alpfa1 9.10...

        i think i chance to 10.04 in 2-3 weeks ........



        U/K9.10+catalyst9.10 is a complete fail:

        Black-skin-bug in wine@Oblivion
        composit->2 sec lag bug by Maxi a programm from the taskbar.
        Kubuntu is unusable lag and some programs like firefox do not work.
        Flash movies in the browsers lag@fullscreen
        KDE4 ---> VESA is faster than Catalyst....


        and yes Positiv part: 3D nativ-games like HON and Chromium work well..

        Comment

        Working...
        X