Hi, I've run both versions on the test page http://www.24fun.com/downloadcenter/benchjs/start.html which is a good start point for this kind of tests I think.
It was tested on a dual 1.8ghz G5 in equal conditions.
New Event Handling Camino: TEST 1 time: 6.2 sec. TEST 2 time: 0.959 sec. TEST 3 time: 1.663 sec. TEST 4 time: 0.869 sec. TEST 5 time: 0.36 sec. TEST 6 time: 3.679 sec. TEST 7 time: 1.103 sec. TOTAL: 14.833 sec.
Old Event Handling Camino: TEST 1 time: 1.678 sec. TEST 2 time: 0.909 sec. TEST 3 time: 1.387 sec. TEST 4 time: 0.843 sec. TEST 5 time: 0.332 sec. TEST 6 time: 3.107 sec. TEST 7 time: 1.059 sec. TOTAL: 9.315 sec.
Note that while the new event handling camino overall performs slower
(mainly bevause of the huge difference in Test 1, which is a Javasript
loop), Test 6 (which is about moving a lot of layers) performs much smoother
than with the OEH - much more like Safari does in this kind
of things. The OEH version seems to "cheat" on that test (6) by just skipping
a lot of frames.
Keeping an eye on top showed only one peak at 100% on test
7 for both versions. Overall, the NEH Camino kept usage under 30% CPU
while the OEH Camino was around 60%.
If I find out the right flags for top to sample Camino's CPU usage in a less
"user-dependant" way, I'll be happy to send you the results... :)
Hope it can help for the beginning,
Regards, Lorenzo
On 21. Nov 2004, at 18:37, Geoff Beier wrote:
Hi All,
In the past we've speculated that the way Camino handles gecko events may be
responsible for some of the various performance issues people report with
Camino. This includes but is not limited to CPU usage during flash playback.
I've got a patch up: https://bugzilla.mozilla.org/show_bug.cgi?id=271050
which drastically changes how gecko events are dispatched. Because this code
is invoked so frequently and the change is so dramatic, I need a lot of
testing before this will sound like a good idea :-).
It would help out a lot if as many folks as possible could either build and
test or download the disk image here:
http://www.speakeasy.org/~gbeier/Camino-perftest.dmg.bz2
and browse with the two builds included there. The one in the "Old event
handling" folder should be very similar to tonight's nightly. The only
difference between that one and the one in the "New event handling" folder
is the above patch; this should make it easily to isolate performance
differences introduced by this change.
Anyone who gets to kick it around, please report results back either to this
list or on the bug itself.
Thanks in advance!
Regards,
Geoff
_______________________________________________ Camino mailing list [EMAIL PROTECTED] http://mozdev.org/mailman/listinfo/camino
_______________________________________________ Camino mailing list [EMAIL PROTECTED] http://mozdev.org/mailman/listinfo/camino
