Bug 20116 - [855GM] xorg-1.5.3 + intel-2.6.1 + linux-2.6.2[89] slow and corrupts display
Summary: [855GM] xorg-1.5.3 + intel-2.6.1 + linux-2.6.2[89] slow and corrupts display
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: unspecified
Hardware: Other All
: medium normal
Assignee: Eric Anholt
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords: NEEDINFO
Depends on:
Blocks:
 
Reported: 2009-02-14 03:48 UTC by Bruno
Modified: 2009-07-22 10:31 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
Screensht of corruption (13.11 KB, image/png)
2009-02-14 03:48 UTC, Bruno
no flags Details
Xorg config (4.52 KB, text/plain)
2009-02-14 03:49 UTC, Bruno
no flags Details
Xorg log (24.71 KB, text/plain)
2009-02-14 03:49 UTC, Bruno
no flags Details
my xorg.conf (310 bytes, text/plain)
2009-03-18 12:51 UTC, Maximilian Engelhardt
no flags Details
my Xorg.0.log (25.54 KB, text/x-log)
2009-03-18 12:52 UTC, Maximilian Engelhardt
no flags Details

Description Bruno 2009-02-14 03:48:30 UTC
Created attachment 22941 [details]
Screensht of corruption

When running on recent kernel (2.6.28 and now 2.6.29-rc5, not verified with earlier kernels) X is slow and CPU-intensive for a lot of operations that worked smoothly with older intel driver, xorg server and kernel.
A sample are the enlightenment animations which previously used nearly-0% CPU time and now use 60% CPU (Xorg) and partially flicker.

In addition the bottom half of the display often suffers from corruption (e.g. partially covered by brown lines/areas) See attached screenshot. The corruption can usually be triggered by GTK comboboxes or menus.

Kernel is running with KMS disabled (KMS not working on my system, see #20115 for system details).
Comment 1 Bruno 2009-02-14 03:49:19 UTC
Created attachment 22942 [details]
Xorg config
Comment 2 Bruno 2009-02-14 03:49:56 UTC
Created attachment 22943 [details]
Xorg log
Comment 3 Thorsten Vollmer 2009-02-14 04:50:29 UTC
The screen corruption looks similar to the one described in bug 18503.
Comment 4 Bruno 2009-02-14 05:05:16 UTC
(In reply to comment #3)
> The screen corruption looks similar to the one described in bug 18503.
They are probably the same, note that I could take the screenshot with gimp without difficult (just a fullscreen screenshot, then extracting pertinent area)
For me corruptions are in range 640..1050px where top of display is at 0px.

So the slowness remains for this bug.
Comment 5 Maximilian Engelhardt 2009-03-18 12:50:28 UTC
Seems to be exactly the same problem that I have with my 855GM card.
I often get screen corruption and X is quite slow.
I'm not 100% sure about this, but I think KDE/Qt is much more affected by the slowdown than gnome/gtk. Perhaps this might be helpful for someone.
Comment 6 Maximilian Engelhardt 2009-03-18 12:51:30 UTC
Created attachment 24006 [details]
my xorg.conf
Comment 7 Maximilian Engelhardt 2009-03-18 12:52:19 UTC
Created attachment 24007 [details]
my Xorg.0.log
Comment 8 Eric Anholt 2009-07-15 14:36:00 UTC
2D performance for software fallbacks (which 8x does a lot of) should be much improved in master or the 2.8 RCs.  Could you retest with that?  kernel 2.6.30 is required, though

Comment 9 Eric Anholt 2009-07-22 10:31:36 UTC
Unfortunately this bug was about two separate issues simultaneously, but I think both are fixed now.

Rémi reports that the corruption is fixed between 2.6.1 and 2.6.2.

The 2.8.0 driver has many sw fallback fixes beyond what 2.6.1 had, and I just pushed a performance fix to master for non-component-alpha text on 8xx on top of some general performance fixes for cairo.  The 8xx render support should now be in better shape on 855 than it's ever been.

commit 12c5aeca7a3db92d3522d00f5daf338d522e2176
Author: Eric Anholt <eric@anholt.net>
Date:   Thu Jul 16 13:34:09 2009 -0700

    8xx render: Add limited support for a8 dests.
    
    This improves aa10text performance from 74k to 569k on my 855 laptop.
    This also causes my 865 to hang on aa10text like it does on rgb10text,
    thanks to actually hitting render accel.


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.