|
Description
|
Problem with the open source ATI driver on Xorg. Mouse responsivness becomes "jerky" with a load on the system. Looks like the ATI driver cannot take advantage of the frame buffer and all the computations needed for mouse (pointer) movement are done in CPU, thus inducing CPU spikes (> 30%) and mouse becomes "jerky."
Cindy,
We need an outlook schedule on this bug. Read the customers response below:
I need an outlook on when I will get a meaningful status on this issue.
There are two product programs carrying this as a critical problem and
all I can say is that I'm waiting for a response from Sun. Does the
Solaris team acknowledge an issue and has it been characterized ? Is it
being worked as a priority and is there a fix outlook ?
It's time for someone to take a stand on this issue. Will you or will you not fix this bug.
Joe
I've been asked to provide a better description of how to reproduce
the issue being reported by this CR.
The symptoms can be reproduced using the Xorg ATI driver on an ATI RageXL
video chip. I've seen the symptoms both on Marrakesh machines using
the on-motherboard ATI RageXL chip, as well as on a non-Marrakesh machine
with a ATI RageXL PCI add-in card. The Xorg driver in both cases is
built from the X.Org Foundation X11R6.8.2 source release. The 6.8.0
driver also demonstrates the same symptoms.
A typical way to reproduce the symptoms under Solaris is as follows:
1. Install Solaris, either 10 or 11 on a machine. S10 RR came with
the 6.8.0 ATI driver. S10U1_13 has the 6.8.2 ATI driver. S11_16
and later uses the 6.8.2 ATI driver.
2. Verify that Xorg is being used as the X server and is configured
for a 24-bit color resolution for the display.
3. Start a CDE session and login.
4. Run prstat in a terminal window.
5. Start a Mozilla browser, and go to http://www.sun.com .
6. There are two areas of the web page that causes dynamic changes
to the web page display:
a) Toward the top right of the page are three fields labeled "A
Hardened Environment", "Talk to the Experts", and "Messaging
for Less". Moving the pointer over these three fields causes
the picture to the left of the fields to change.
b) Underneath the picture, there are six headings: "Products",
Downloads", ... "Research". Moving the pointer over these
headings causes a menu to drop down.
7. Rapidly move the cursor over all of these areas and watch the
CPU utilization of Xorg in the prstat window.
8. Two things should be noticed:
a) Xorg CPU utilization will climb very quickly to over 50% and
eventually over 80%.
b) The dynamic changes to the screen are sluggish, and cause
the on-screen pointer to lag the movement of the pointing
device.
On other (more current) video cards, the dynamic screen changes are
much snappier, and there is no noticable lag between the on-screen
pointer and the pointing device movement. Although the CPU utilization
can be driver up to 40% or more, it takes much longer to do so.
It has been reported that on a Linux installation, the behavior is
the same with the Xorg ATI driver, but is greatly improved when using
the proprietary ATI driver that is available from the ATI web site.
The proprietary driver can be obtained by going to http://www.ati.com,
selecting "Drivers & Software", "Linux Drivers and Software", and then
"Motherboards with ATI Graphics". The current version is 8.16.20.
Edward Shu in Beijing was going to try to verify this.
I'll also note that this is not an issue of hardware vs software
cursor. Although the use of the software cursor might show similar
symptoms, I've been able to determine that the hardware cursor
feature is enabled for this driver.
With the new drivers for 6.8.2, the performance is significantly improved. The output no longer lags the mouse in the sun web site. If you work at it for a long time, you can still get up to 70-75% CPU utilization for xorg on the prstat screen, but it takes 3-4 minutes as opposed to less than 1 minute. The user no longer sees a problem as they move the mouse.
Also checked with 6.9rc1 drivers. Again the delay went from very obious to barely perceivable with similar prstat results.
the high usage of CPU is normal. I noticed that even in MS windows the CPU usage can rise to
60% if we repeat the operation mentioned above.
Per email from Cindy not a s10u2 Beta stopper
|
