Comment #32 on issue 7823 by [email protected]: REGRESSION: Closing tabs  
takes too long
http://code.google.com/p/chromium/issues/detail?id=7823

@divilex
Couldn't reproduce in r22886 from build.chromium.org.  What version are you  
using?

@pkasting
(this comment started before you posted 31)
What sites are you seeing the slowdown on besides shacknews.com?

r22818 was recent and plugin-related, maybe a negative side-effect worth  
investigating is
present.  In certain cases (last tab with a plugin reference) reloads  
should be
noticeably faster.  In those cases I expect tab closing to be faster, as  
well.

I ran a series of tests using the stopwatch on my phone to gauge how long  
it took to
close tabs with shacknews.com and the laryn.x?story=59903 URLs.  I could  
only induce a
very-slow-to-close state when my copies of Chromium (2.0 and r22886) were  
loaded down with many tabs (20 each).  My tests wouldn't highlight a single  
delay of 20ms given my
measuring instrument, but it was my hope any delay would be replicated per  
tab across 10
or 20 tabs.  Under the generally loaded state, tab closing at times  
felt "sluggish" and
at other times felt "snappy."

In one test (first closing Chrome 3.0 with 20 tabs then closing Chrome 2.0  
with 20 tabs),
there was a hang of >15s in closing a tab in Chrome 2.0.  I ran out of time  
before I
could repeat this test.

Anyway, here are the results of r22886 compared to 2.0.172.39 using my  
phone's stopwatch.
My laptop runs Windows XP on an Intel L7500 @ 1.6GHz, 1.96 GB of RAM.  I  
was concerned
that given 2 GB of RAM, it was important in a field of 2 builds (a, b) to  
run the tests
in order (a, b), then again in order (b, a) to give both a chance to  
operate under
greater and lesser loads.  If someone were to run these again, I'd suggest  
to also note
if closing felt sluggish or snappy.  I regret not doing so here.

=== Test series 1
Start with r22886 (3.0.198.0) and 2.0.172.39, each with 10 shacknews.com  
tabs + 1
about:memory tab.  about:memory reported each tab had its own renderer  
process.

Test 1: closing all tabs left to right in r22886, then all tabs left to  
right in
2.0.172.39.
#  mem(cr3)  mem(cr2)  t(cr3)   t(cr2)
0  580 MB    435 MB    15.9s    7.5s
1  602 MB    448 MB    6.0s     6.5s
2  572 MB    415 MB    6.2s     6.3s

Test 2: closing all tabs left to right in 2.0.172.39, then all tabs left to  
right in
r22886.
#  mem(cr2)  mem(cr3)  t(cr2)   t(cr3)
0  396 MB    643 MB    15.6s    5.4s*
1  459 MB    587 MB    6.4s     5.2s
2  413 MB    564 MB    5.2s     5.1s
* test 2.0 run right after test 1.0

Test 3: closing all tabs right to left in r22886, then all tabs right to  
left in
2.0.172.39.
#  mem(cr3)  mem(cr2)  t(cr3)   t(cr2)
0  538 MB    407 MB    6.5s     5.9s

Test 4: closing all tabs right to left in 2.0.172.39, then all tabs right  
to left in
r22886.
#  mem(cr2)  mem(cr3)  t(cr2)   t(cr3)
0  419MB     568 MB    6.8s     6.2s

=== Test series 2
Start with 3.0.198.0 and 2.0.172.39, each with 20 shacknews.com tabs + 1  
about:memory
tab.  about:memory reported each tab in r22886 had its own renderer,  
2.0.172.39 sometimes
reused renderer processes.

Test 5: closing all tabs left to right in r22886, then all tabs left to  
right in
2.0.172.39.
#  mem(cr3)  mem(cr2)  t(cr3)   t(cr2)
0  130 MB    167 MB    37.0s    57.8s*
* significant hang >15s

Test 6: closing all tabs left to right in 2.0.172.39, then all tabs left to  
right in
r22886.
#  mem(cr2)  mem(cr3)  t(cr2)   t(cr3)
0  217 MB    348 MB    42.7s    32.9s

=== Test series 3: load http://www.shacknews.com/laryn.x?story=59903 in 10  
tabs + 1
about:memory tab

Test 7: closing all tabs left to right in r22886, then all tabs left to  
right in
2.0.172.39.
#  mem(cr3)  mem(cr2)  t(cr3)   t(cr2)
0  326 MB    292 MB    5.0s     4.7s

Test 8: closing all tabs left to right in 2.0.172.39, then all tabs left to  
right in
r22886.
#  mem(cr2)  mem(cr3)  t(cr2)   t(cr3)
0  256 MB    248 MB    5.6s     4.8s

=== Conclusions
When my laptop is overloaded with shacknews.com tabs, both versions have  
noticeable
sluggishness while closing those tabs.  10 tabs early on induced a  
slow-to-close-tabs
state, similar tests later were snappy.  20 tabs induced a  
very-slow-to-close-tabs state.

In tests 1.0 and 2.0, both versions showed inconsistency under high load,  
this might be
attributed to my laptop settling in as I began this test series.  I could  
not repeat this
condition in the later runs of test 1 or 2.

In test 5.0, 2.0.172.39 had a noticeable hang that severely affected its  
results.

In tests 5.0 and 6.0, r22886 was sluggish yet had consistent performance  
under greater
and lesser loads.

There is a general trend across the tests that the results for each run  
were faster than
previous tests results.  I attribute this to my getting faster at running  
the tests over
time and not to a difference between Chromium builds or test parameters.

--
You received this message because you are listed in the owner
or CC fields of this issue, or because you starred this issue.
You may adjust your issue notification preferences at:
http://code.google.com/hosting/settings

--~--~---------~--~----~------------~-------~--~----~
Automated mail from issue updates at http://crbug.com/
Subscription options: http://groups.google.com/group/chromium-bugs
-~----------~----~----~----~------~----~------~--~---

Reply via email to