[Bug 38800] glXSwapBuffersMscOML is slow on AMD Fusion but not on Intel 945 w/Atom

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Tue Jul 5 10:57:45 PDT 2011


https://bugs.freedesktop.org/show_bug.cgi?id=38800

--- Comment #17 from Simon Farnsworth <simon.farnsworth at onelan.co.uk> 2011-07-05 10:57:45 PDT ---
I added "video=LVDS-1:d" to my kernel command line, which changes xrandr output
to:

# xrandr 
Screen 0: minimum 320 x 200, current 1920 x 1200, maximum 8192 x 8192
LVDS disconnected (normal left inverted right x axis y axis)
HDMI-0 connected 1920x1200+0+0 (normal left inverted right x axis y axis) 519mm
x 324mm
   1920x1200      60.0*+
   1920x1080      60.0  
   1600x1200      60.0  
   1280x1024      75.0     60.0  
   1152x864       75.0  
   1024x768       75.1     60.0  
   800x600        75.0     60.3  
   640x480        75.0     60.0  
   720x400        70.1  
VGA-0 disconnected (normal left inverted right x axis y axis)

It also stops the HDMI interface working; I have no picture once KMS kicks in.

Frame rate does not improve; r600_fence_finish is now around 52% of my CPU
load, while delay_tsc from the kernel has moved up to be around 5%, replacing
radeon_fp_native_mode.

There has been no change in overall CPU usage - I'm still seeing the same 27%
used by my test app, 80% overall idle time.

I'm going home now (it's nearly 7pm here) - if there are any tests I should try
for you tomorrow, please let me know; getting the Fusion up to speed compared
to Intel is my current work priority. I'm happy to try random git trees,
arbitrary patches, or just editing code as described by someone else.

I have confirmed that the Intel drivers I'm using have a working glFinish()
implementation; rendering always completes before swapbuffers; in as far as
possible, I'm comparing like-with-like, and the baseline for both systems is
Fedora 15.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.


More information about the dri-devel mailing list