Bug 14161 - [830M EXA] EXA causes Xorg to take up too much cpu time
Summary: [830M EXA] EXA causes Xorg to take up too much cpu time
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: unspecified
Hardware: x86 (IA32) Linux (All)
: medium major
Assignee: Eric Anholt
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords: NEEDINFO
Depends on:
Blocks: intel-2.5
  Show dependency treegraph
 
Reported: 2008-01-20 15:42 UTC by Lauri Võsandi
Modified: 2008-09-09 06:50 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
Xorg.0.log: Forced XAA, Intel video driver version 2.2.0, working properly (34.49 KB, text/plain)
2008-01-23 05:58 UTC, Lauri Võsandi
no flags Details
xorg.conf: Forced XAA; Intel video driver 2.2.0; working properly (1.85 KB, application/octet-stream)
2008-01-23 05:59 UTC, Lauri Võsandi
no flags Details

Description Lauri Võsandi 2008-01-20 15:42:40 UTC
This happens in xf86-video-intel-2.1.1-2, whereas it didnt in xf86-video-intel-2.2.0-1. When I downgraded my video driver to older version, everything works much better. This happens on ArchLinux and. I observed similar behaviour in Ubuntu also, after driver upgrade things got sluggish. Latest version causes Xorg to take too much cpu time, making computer almost useless. This happens on IBM X30 (using I810 chipset). Looks as if writing buffers from RAM to videocard's memory is somehow buggy or slow. Doesnt cause lockups, just X11 based graphic-intensive applications are very-very slow (Firefox on a page with many pictures), at the same time XV based applications run fine (movie players etc). I can test and give you guys feedback. Lets get this thing fixed!
Comment 1 Gordon Jin 2008-01-20 17:13:39 UTC
Please attach xorg.conf and Xorg.0.log.

Does it work better with 2.2.0 if you set XAA in xorg.conf?
Comment 2 Lauri Võsandi 2008-01-20 23:14:18 UTC
(In reply to comment #1)
> Please attach xorg.conf and Xorg.0.log.
> 
> Does it work better with 2.2.0 if you set XAA in xorg.conf?
> 

First I messed up first sentence: 2.2.0 doesnt work, but 2.1.1 works. I see 2.2.0 defaults to EXA. When I switched from EXA to XAA by  Option "AccelMethod" "XAA", everything was back to normal. I see you guys are writing new acceleration methods for Xorg, but I'd suggest to use XAA by default on older chipsets as i810, since none wont possibly be running compositing manager on such old piece of hardware anyway.
Comment 3 Lauri Võsandi 2008-01-21 09:21:55 UTC
It seems that 2.2.0 driver crashes a lot when switched back to XAA instead of EXA. Especially when using vncviewer. 
Comment 4 Gordon Jin 2008-01-21 18:41:51 UTC
log please. I wonder if it's 810 or 830.
Comment 5 Lauri Võsandi 2008-01-23 05:58:01 UTC
Created attachment 13876 [details]
Xorg.0.log: Forced XAA, Intel video driver version 2.2.0, working properly
Comment 6 Lauri Võsandi 2008-01-23 05:59:46 UTC
Created attachment 13877 [details]
xorg.conf: Forced XAA; Intel video driver 2.2.0; working properly
Comment 7 Lauri Võsandi 2008-01-23 14:39:21 UTC
(==) Configured Mouse: YAxisMapping: buttons 4 and 5
(**) Configured Mouse: EmulateWheel, EmulateWheelButton: 2, EmulateWheelInertia: 10, EmulateWheelTimeout: 200
(**) Configured Mouse: Buttons: 9
(**) Configured Mouse: Sensitivity: 1
(II) evaluating device (Configured Mouse)
(II) XINPUT: Adding extended input device "Configured Mouse" (type: MOUSE)
(II) evaluating device (Generic Keyboard)
(II) XINPUT: Adding extended input device "Generic Keyboard" (type: KEYBOARD)
(II) Configured Mouse: ps2EnableDataReporting: succeeded
AUDIT: Wed Jan 23 23:56:16 2008: 6705 X: client 23 rejected from local host (uid 0)
AUDIT: Thu Jan 24 00:12:00 2008: 6705 X: client 24 rejected from local host (uid 0)

Backtrace:
0: /usr/bin/X(xf86SigHandler+0x7e) [0x80bddee]
1: [0xb7f0d420]
2: /usr/lib/xorg/modules/drivers//intel_drv.so [0xb7b1c827]
3: /usr/bin/X [0x80de315]
4: /usr/lib/xorg/modules/extensions//libextmod.so(XvdiPutImage+0x178) [0xb7b686d8]
5: /usr/lib/xorg/modules/extensions//libextmod.so [0xb7b6b546]
6: /usr/bin/X [0x814b98e]
7: /usr/bin/X(Dispatch+0x2bf) [0x808c9cf]
8: /usr/bin/X(main+0x48b) [0x8073f3b]
9: /lib/libc.so.6(__libc_start_main+0xe0) [0xb7cd6390]
10: /usr/bin/X(FontFileCompleteXLFD+0x209) [0x80732b1]

Fatal server error:
Caught signal 11.  Server aborting

(II) AIGLX: Suspending AIGLX clients for VT switch


VNC and XV cause such errors on 2.2.0 with XAA
Comment 8 Gordon Jin 2008-01-23 17:16:30 UTC
So it's 830M, not 810.

Let's focus on EXA issue on this bug.
Comment 9 Lauri Võsandi 2008-01-24 13:11:59 UTC
XAA will be obsolete soon? What can I do about EXA?

Using XAA (2.2.0) minimized mplayer (XV outpu) and restored again and then I got another crash:
Backtrace:
0: /usr/bin/X(xf86SigHandler+0x7e) [0x80bddee]
1: [0xb7f7b420]
2: /lib/libc.so.6(vsnprintf+0xb4) [0xb7d8ec84]
3: /usr/bin/X(LogVWrite+0xb7) [0x81b8e47]
4: /usr/bin/X(LogVMessageVerb+0x99) [0x81b9379]
5: /usr/bin/X(xf86VDrvMsgVerb+0xda) [0x80cab6a]
6: /usr/bin/X(xf86DrvMsg+0x3d) [0x80cbb4d]
7: /usr/lib/xorg/modules/drivers//intel_drv.so [0xb7b7fc81]
8: /usr/lib/xorg/modules/drivers//intel_drv.so(i830_free_memory+0x2e) [0xb7b7fd19]
9: /usr/lib/xorg/modules/drivers//intel_drv.so [0xb7b8590b]
10: /usr/bin/X [0x80dc9cc]
11: /usr/lib/xorg/modules/extensions//libextmod.so [0xb7bd5dfa]
12: /usr/bin/X [0x80dcdf8]
13: /usr/lib/xorg/modules/extensions//libdbe.so [0xb7f6e865]
14: /usr/bin/X(compDestroyWindow+0xc5) [0x8101d65]
15: /usr/bin/X [0x807b3c1]
16: /usr/bin/X(DeleteWindow+0x158) [0x807b5b8]
17: /usr/bin/X(FreeResource+0x119) [0x8075ab9]
18: /usr/bin/X(ProcDestroyWindow+0x7f) [0x8088c3f]
19: /usr/bin/X [0x814ba0e]
20: /usr/bin/X(Dispatch+0x2bf) [0x808c9cf]
21: /usr/bin/X(main+0x48b) [0x8073f3b]
22: /lib/libc.so.6(__libc_start_main+0xe0) [0xb7d44390]
23: /usr/bin/X(FontFileCompleteXLFD+0x209) [0x80732b1]
Comment 10 Wang Zhenyu 2008-03-13 19:49:59 UTC
Lauri, Xv crash bug has been fixed in current git master. Could you try it?

For exa performance issue, we have seperate bug to track, and the work is under way in intel-batchbuffer branch.
Comment 11 Eric Anholt 2008-04-18 01:27:30 UTC
I fixed a major performance issue with EXA on 830 machines a month or two ago.
Comment 12 Michael Fu 2008-07-24 20:21:26 UTC
Lauri, ping for response...
Comment 13 Gordon Jin 2008-09-09 06:50:02 UTC
feedback timeout, so closing.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.