Xorg 7 uses much more CPU than Xorg 6

Immediately after upgrading from Xorg 6 to 7, I noticed that
whenever an 
xterm is "busy" (e.g. compiling something), the
entire desktop is sluggish. 
  I've noticed this on at least three machines with vastly
differing hardware.

To quantify my experience, I created to slices on one i386
machine and 
installed two systems:
     FreeBSD 6.2 with X.org 6.9
     FreeBSD 6.3 with X.org 7.3
I created the initial xorg.conf file with "X
-configure" and made trivial 
modifications for my mouse and monitor.

Running only twm and some xterms, I compiled the 6.2-RELEASE
world on each 
system and recorded the CPU time used by the Xorg server:
     Xorg 6   7.23 seconds
     Xorg 7  17.77 seconds

I also ran "yes" with an 80-character argument and
ran top to watch CPU 
usage.  After top reached a steady state, I recorded the
following:

==== Xorg 6 ====

CPU states: 67.3% user,  0.0% nice, 32.7% system,  0.0%
interrupt,  0.0% idle

   PID USERNAME  THR PRI NICE   SIZE    RES STATE    TIME  
WCPU COMMAND
   709 root        1 120    0  5240K  4200K RUN      1:08
61.24% xterm
   702 root        1 107    0 54636K 10176K select   0:47
30.77% Xorg
   729 root        1   6    0  1260K   616K ttywri   0:04 
3.28% yes

==== Xorg 7 ====

CPU states: 25.9% user,  0.0% nice, 74.1% system,  0.0%
interrupt,  0.0% idle

   PID USERNAME    THR PRI NICE   SIZE    RES STATE    TIME 
 WCPU COMMAND
   768 root          1 122    0 58408K 13188K RUN      1:13
71.44% Xorg
   777 root          1 106    0  5344K  4284K RUN      0:25
24.77% xterm
   793 root          1   6    0  1268K   604K ttywri   0:02 
0.79% yes

================

Does anyone have any ideas why this is happening?  I'll
gladly supply my 
configurations, but they are all remarkably generic.

Thanks,

Eric
_______________________________________________
freebsd-x11freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-x11
To unsubscribe, send any mail to
"freebsd-x11-unsubscribefreebsd.org"

On Mon, Apr 14, 2008 at 09:50:20PM -0500, Eric van Gyzen
wrote:
>I also ran "yes" with an 80-character argument
and ran top to watch CPU 
>usage.  After top reached a steady state, I recorded the
following:

Note that it's very difficult to see differences in
performance with
this test - in either case, the system is flat out but
there's no
indication of how fast it's processing.  You might like to
experiment
with ports/x11/x11perf

>==== Xorg 6 ====
>CPU states: 67.3% user,  0.0% nice, 32.7% system,  0.0%
interrupt,  0.0% idle
>
>==== Xorg 7 ====
>CPU states: 25.9% user,  0.0% nice, 74.1% system,  0.0%
interrupt,  0.0% idle

The most striking thing to me is the difference in system
time.  You
might like to ktrace both Xorg's whilst doing some fixed
workload and
compare the system calls.  There was a recent thread about
Xorg doing
lots of gettimeofday(2) syscalls due to a configure error
and possibly
you are hitting this or a related issue.

-- 
Peter Jeremy
Please excuse any delays as the result of my ISP's inability
to implement
an MTA that is either RFC2821-compliant or matches their
claimed behaviour.