high cpu load on http://www.kranten.com

[vln]

Binary package hint: firefox

Site: http://www.kranten.com

Problem: When this particular site is opened the CPU load is constantly at 40-90%

Affected versions: Ubuntu's Firefox 3

[vln (ehitajate746)]

Firefox 3 beta 2 has also the same problem

[Matthew Woerly (nattgew)]

I can confirm. Do all sites with Java animation do this? I have seemed to find that regardless of the complexity of the animation, it will use most of available processor, yet anything from simple things like this to RuneScape or other complicated interfaces also work smoothly.

[Matthew Woerly (nattgew)]

2.0.0.11 here too, what Java do you have installed?

[vln (ehitajate746)]

java --version:

Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0_13-b05)

[vln (ehitajate746)]

tried it also with jre build 1.6.0_03-b05, problem remains

[Pjotr12345 (computertip)]

+1

Bad bug. I can't play chess anymore on Yahoo Games, because then the CPU usage on my Intel Core 2 Duo machine skyrockets.

[Matti Lindell (mlind)]

News ticker on that site is made using JavaScript and is not Java related. The bug is either in firefox' JavaScript engine or in the script itself. I'll close the Java related tasks.

[Matthew Woerly (nattgew)]

Due to this comment https://bugs.launchpad.net/ubuntu/+source/firefox-3.0/+bug/184740/comments/9 I think that this one would be in the same category...

[Pjotr12345 (computertip)]

I don't have the problem anymore on Yahoo Games. But I still do have it on a website that uses a news ticker:
http://www.kranten.com

So I think that the conclusion of mlind is correct: the problem lies either in Firefox' JavaScript engine or in the script itself.

[Alexander Sack (asac)]

firefox 2 wont receive any fixes anymore.

[Alexander Sack (asac)]

the updated url in summary triggers the high cpu load for me.

[Alexander Sack (asac)]

this bug is in xulrunner 1.9.

[Pjotr12345 (computertip)]

This bug is still present in Firefox 3.0 in Ubuntu 8.04 Hardy Heron alpha 5.
Check http://www.kranten.com (the news ticker is the culprit).

[Daniele Dellafiore (ildella)]

Here ff3b3 on hardy up to date is very slow in switching between tabs.
Took more than a second, whatever the content of the tabs is.

FF3 is much faster on gutsy on my three years old desktop than in my new laptop. Very slow in opening a new tab also.

[Daniele Dellafiore (ildella)]

ok I think here we are facing a real nvidia driver problem.
You can read my various test on my two system on bug 144216 comments.

I can both experience the slow tab switching problem and also some strongly javascript based sites (like www.netvibes.com that I use everyday) are *really* slow with nvidia+compiz, while are really fast without them.

I think I will point these two bugs to nvidia tomorrow. You think compiz team is aware about this?

This problems definitely does not allow me to use compiz and friends on my laptop and the nvidia driver as well. I need them to develop application...

And Dell, too, they sell this systems with Ubuntu 7.10, they should care about this problem...

[Pjotr12345 (computertip)]

@della: the bug has *nothing* to do with Nvidia. I have the bug on two systems with Intel video cards. It's a bug in Firefox itself, in the way it handles certain java scripts (news tickers). :-)

Greetz, Pjotr.

[MaitoX (quimariano)]

I can confirm what Pjotr12345 says, I even see ff3 b5 eating up to 50% of my CPU under Windows XP when visiting that site, so it's a FF bug, and it's cross platform.

[simtris (simtris)]

Same issue about me :
FF3 but same for FF2
Firefox is running very nice on windows but is a little slow on ubuntu. I have an NVIDIA card and compiz activated.
My own website isn't very fluid : http://www.ubuntu-dijon.org/librexpo/
There bust be a smooth fading beteween each pages and on my Ubuntu it's just appear brutally.

Don't know why.

I have Hardy but have the same issue on Gusty

[Kari Aliranta (kpaliran)]

Here's a benchmark originally created for another Firefox bug, but probably usable also here, since it uses javascript:

http://launchpadlibrarian.net/1585880/test.html

Personally, Firefox 3rc1 on Hardy took about 6 times more time to complete all tests (not only text rendering) than Firefox 2 on Gutsy. So I'd say confirmed also for Xulrunner used in rc1.

[Kari Aliranta (kpaliran)]

Sorry, I'll take back my last comment. The javascript part of the test is actually 3 times faster in FF3 than on FF2, and the other tests are about the same on both versions on Hardy - also FF3 is about 4 times faster in Sunspider benchmark than FF2. HOWEVER, FF2 on Gutsy was about six times faster in the page rendering tests than either version on Hardy, and also the page rendering is visually noticeably slower on Hardy. There is something funny going on...

[simtris (simtris)]

So...
Somthing around Hardy ? Quite possible. Anyway I tried some fixes during this time and I discover an French thread abaout this kind of problem. The advise was to modify the /etc/hosts, this way :

127.0.0.1 localhost buntuStation
#127.0.1.1 buntuStation

The second line was the original one and seems to decresse something in Ubuntu/FF.
My Firefox is much quick after that but, on Gmail I always have a High CPU load and a slow scrolling.
...
Didn't understand.
I'll soon try and complete' from scratch, reinstallation of the 8.04, maybe the problem will be fixed ...

 Simon

_________________________________________________________________
Votre contact a choisi Hotmail, l'e-mail ultra sécurisé. Créez un compte gratuitement !
http://www.windowslive.fr/hotmail/default.asp

[Alexander Sack (asac)]

On Thu, Jun 12, 2008 at 11:24:27PM -0000, Kari Aliranta wrote:
> Sorry, I'll take back my last comment. The javascript part of the test
> is actually 3 times faster in FF3 than on FF2, and the other tests are
> about the same on both versions on Hardy - also FF3 is about 4 times
> faster in Sunspider benchmark than FF2. HOWEVER, FF2 on Gutsy was about
> six times faster in the page rendering tests than either version on
> Hardy, and also the page rendering is visually noticeably slower on
> Hardy. There is something funny going on...
>

Which cairo package version are you using? Are you using XAA or EXA
accelmethod in X? Maybe attach your Xorg.0.log and xorg.conf

 - Alexander

[Kari Aliranta (kpaliran)]

>Which cairo package version are you using? Are you using XAA or EXA
>accelmethod in X? Maybe attach your Xorg.0.log and xorg.conf

Alexander: Ok, I checked my notes, and here's some pretty detailed and a bit corrected information (sorry for two posts, there are two attachments and I only use the web interface for Launchpad):

The attachments are my current (testcase #2 and #3 below) Xorg.o.log and xorg.conf on Xubuntu HARDY, with Cairo version 1.6.0. The accelmethod in all cases is XAA, of course.

*** #1 ***
The abovementioned benchmark and Gtkperf results with FF 2.0.0.14 on GUTSY, with Pango disabled (to avoid https://launchpad.net/bugs/32561 ). The computer was a (now sold) 1.1Ghz Athlon with Nvidia Geforce 2 MX 400 and using the nv driver, with an unmodified xorg.conf - except for the custom modeline that I also use on my never computer below (removing it and using the standard modes doesn't help). The results:

Javascript interpreter benchmark: not tested, I hunted for another bug back then.
Scrolling benchmark w/ text: about 2.5 seconds
Scrolling benchmark w/ divs: < 2.5 seconds
Scrolling benchmark w/ nothing: < 2.5 seconds

Gtkperf result with 100 test rounds: about 26 seconds.
General feel when scrolling long pages with lots of text and pictures: very snappy.

*** #2 ***
The abovementioned benchmark and Gtkperf results with FF3 rc1 on HARDY, with Pango disabled (enabling it doesn't seem to have significant effect on FF3) and without any extensions. The computer is a 2.4Ghz P4 with Nvidia Geforce 2 MX 400, using the nv driver:

Javascript interpreter benchmark: about 5 seconds
Scrolling benchmark w/ text: about 14 seconds
Scrolling benchmark w/ divs: about 7 seconds
Scrolling benchmark w/ nothing: about 5 seconds

Gtkperf result with 100 test rounds: about 19 seconds.
General feel when scrolling long pages with lots of text and pictures: slightly jerky.

*** #3 ***
The abovementioned benchmark with FF 2.0.0.14 on HARDY, with Pango disabled and without any extensions. The computer is the same as in the previous test case, a 2.4Ghz P4 with Nvidia Geforce 2 MX 400, using the nv driver:

Javascript interpreter benchmark: about 12 seconds
Scrolling benchmark w/ text: about 12 seconds (16 seconds with Pango enabled)
Scrolling benchmark w/ divs: about 6 seconds
Scrolling benchmark w/ nothing: about 4 seconds

General feel when scrolling long pages with lots of text and pictures: slightly jerky.

*** #4 ***

I also have less complete notes on making the benchmark with FF3 beta 5 on HARDY, without any extensions and Pango disabled. The computer was a (now also sold) 1.0Ghz Athlon with Nvidia Riva TNT2 64 with nv driver and an unmodified xorg.conf:

Scrolling benchmark w/ text: 25 seconds

General feel when scrolling long pages with lots of text and pictures: annoyingly jerky.

*** Other notes: ***

- All tests are made with Human-theme. Changing the theme to something simple, Mist for example, doesn't affect the results of FF2 or FF3 benchmarks - it, however, significantly lowers the time needed for Gtkperf test.

- Gtkperf result (100 test rounds) with testcases #2 and #3 computer was about 16 seconds on GUTSY and is about 19 se...

[Kari Aliranta (kpaliran)]

Also, I forgot to mention:

- Using the Nvidia proprietary driver doesn't help

- I also tried using pure Vesa driver on my current computer - didn't help either (didn't make the things any worse, though)

- The Cairo version on GUTSY test cases was the standard 1.4.10 form Gutsy repos

Kari

[Alexander Sack (asac)]

On Sun, Jun 15, 2008 at 09:41:27PM -0000, Kari Aliranta wrote:
> The abovementioned benchmark and Gtkperf results with FF 2.0.0.14 on GUTSY, with Pango disabled (to avoid https://launchpad.net/bugs/32561 ). The computer was a (now sold) 1.1Ghz Athlon with Nvidia Geforce 2 MX 400 and using the nv driver, with an unmodified xorg.conf - except for the custom modeline that I also use on my never computer below (removing it and using the standard modes doesn't help). The results:

pango cannot be disabled in ffox 3 anymore afaict.

 - Alexander

[Alexander Sack (asac)]

On Sun, Jun 15, 2008 at 09:48:35PM -0000, Kari Aliranta wrote:
> Also, I forgot to mention:
>
> - Using the Nvidia proprietary driver doesn't help
>
> - I also tried using pure Vesa driver on my current computer - didn't
> help either (didn't make the things any worse, though)
>
> - The Cairo version on GUTSY test cases was the standard 1.4.10 form
> Gutsy repos
>

We have cairo 1.6 not 1.4. what package are you running exactly?

- Alexander

[Kari Aliranta (kpaliran)]

>We have cairo 1.6 not 1.4. what package are you running exactly?

Now, on Hardy, I run Cairo version 1.6.0 (libcairo2 version 1.6.0-0ubuntu1)

On Gutsy I was running version 1.4.10 (libcairo2 version 1.4.10-1ubuntu4.4) as I told above.

>pango cannot be disabled in ffox 3 anymore afaict.

Well, that explains why putting MOZ_DISABLE_PANGO=1 in /etc/environment didn't have any effect on FF3, :).

Kari

[Kari Aliranta (kpaliran)]

I tested removing xulrunner 1.9 altogether and using xulrunner 1.8 instead (with FF2): didn't help, FF still scrolls a lot slower than in Gutsy --> would definitely point to a Cairo problem. I'll try to downgrade Cairo to Gutsy version (1.4.10) and see what happens.

Kari

[Kari Aliranta (kpaliran)]

Sorry, no can do:

My Xubuntu Hardy system had no problem with using Gutsy version of libcairo2 (didn't fix the Firefox problem either), but the Hardy version of Xserver wouldn't even start with the Gutsy version of libpixman (Well, I guess the Xorg guys were right when setting their dependencies...)

If I've understood right, Cairo also uses pixman for software fallbacks, so in this case the problem could also be there...however, as the Nvidia proprietary driver (that supposedly doesn't use much software fallbacks) isn't any faster than nv with Firefox, and the Cairo-powered rendering of Xfce GTK+-user interface is not slower, I very severely doubt that.

Kari

[Alexander Sack (asac)]

On Thu, Jun 19, 2008 at 01:40:48AM -0000, Kari Aliranta wrote:
> Sorry, no can do:
>
> My Xubuntu Hardy system had no problem with using Gutsy version of
> libcairo2 (didn't fix the Firefox problem either), but the Hardy version
> of Xserver wouldn't even start with the Gutsy version of libpixman
> (Well, I guess the Xorg guys were right when setting their
> dependencies...)
>
> If I've understood right, Cairo also uses pixman for software fallbacks,
> so in this case the problem could also be there...however, as the Nvidia
> proprietary driver (that supposedly doesn't use much software fallbacks)
> isn't any faster than nv with Firefox, and the Cairo-powered rendering
> of Xfce GTK+-user interface is not slower, I very severely doubt that.
>

Most likely the free nv is faster as you can use EXA AccellMethod
... give it a try.

 - Alexander

[Kari Aliranta (kpaliran)]

>Most likely the free nv is faster as you can use EXA AccellMethod
>... give it a try.

Sorry, but as the nv driver man page states, EXA is only supported for G80 (Geforce 8xxx) and higher - with my card (Geforce 2), the driver just ignores the option: "(WW) NV(0): Option "AccelMethod" is not used"

Besides, I remind that everything else than Firefox scrolling works well - and the 500%-1000% slowdown compared to Gutsy isn't that bad either on a faster machine, as scrolling was very snappy in Gutsy. But on a computer like low end Pentium III (a good target for Xubuntu) the difference begins to be really noticeable. I would also imagine that the initial page rendering speed is affected, too.

So, for the sake of Xubuntu goals it would be nice to know whether the problem is specific to my two systems or is more general...

Kari

[Kari Aliranta (kpaliran)]

Well, I was wild and installed Nouveau driver for my Nvidia card to get EXA working. The result: about 15% improvement in the benchmark for FF3, but I get hugely improved (>2X) results in Gtkperf text scrolling benchmark and everything else is definitely faster, too. Very nice, but I wouldn't consider this a solution for the FF problem, though.

Kari

[Kari Aliranta (kpaliran)]

  New information:

Using top, I noticed that most of the CPU usage during scrolling is actually due to X, not Firefox process itself - X uses 60-90% of CPU time, Firefox the rest. This may be normal or not, but anyway there is another bug that has more information on that: https://bugs.launchpad.net/firefox/+bug/38131 .

Kari

[simtris (simtris)]

Same issue, same effect here :)Ill réinstall everithing this evening. I think it ll solve that.Simon
_________________________________________________________________
Caroline vient de mettre à jour son profil Messenger ! Connectez-vous !
http://login.live.com/login.srf?wa=wsignin1.0&rpsnv=10&ct=1198837564&rver=4.0.1534.0&wp=MBI&wreply=http:%2F%2Fhome.services.spaces.live.com%2F&lc=1036&id=73625

[Sambit Bikas Pal (sambitbikaspal)]

Merely "reinstalling everything" is not at all a solution.

On Sun, Jun 22, 2008 at 9:58 PM, simtris <email address hidden> wrote:

> Same issue, same effect here :)Ill réinstall everithing this evening. I
> think it ll solve that.Simon
> _________________________________________________________________
> Caroline vient de mettre à jour son profil Messenger ! Connectez-vous !
>
> http://login.live.com/login.srf?wa=wsignin1.0&rpsnv=10&ct=1198837564&rver=4.0.1534.0&wp=MBI&wreply=http:%2F%2Fhome.services.spaces.live.com%2F&lc=1036&id=73625
>
> ** Attachment added: "unnamed"
> http://launchpadlibrarian.net/15512504/unnamed
>
> --
> high cpu load because of JavaScript
> https://bugs.launchpad.net/bugs/185469
> You received this bug notification because you are a member of Ubuntu
> Bugs, which is subscribed to Ubuntu.
>
> --
> ubuntu-bugs mailing list
> <email address hidden>
> https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
>

--
Sambit Bikas Pal
MS 3rd Year
Indian Institute Of Science Education & Research Kolkata,
HC 7, Sector-III
Salt Lake,
Kolkata-700106

Web http://sambitbikaspal.googlepages.com
OpenPGP Key:
http://keyserver.ubuntu.com:11371/pks/lookup?op=get&search=0x8E57F8B897D372B3

[simtris (simtris)]

So, what's the solution ?
_________________________________________________________________
Caroline vient de mettre à jour son profil Messenger ! Connectez-vous !
http://login.live.com/login.srf?wa=wsignin1.0&rpsnv=10&ct=1198837564&rver=4.0.1534.0&wp=MBI&wreply=http:%2F%2Fhome.services.spaces.live.com%2F&lc=1036&id=73625

[Robert (robrwo)]

I've been experimenting with Fedora 9 and note that the same issue
occurs with Firefox 3.

[Alexander Sack (asac)]

On Thu, Jun 19, 2008 at 11:55:03PM -0000, Kari Aliranta wrote:
>
> Well, I was wild and installed Nouveau driver for my Nvidia card to get EXA working. The result: about 15% improvement in the benchmark for FF3, but I get hugely improved (>2X) results in Gtkperf text scrolling benchmark and everything else is definitely faster, too. Very nice, but I wouldn't consider this a solution for the FF problem, though.
>
> Kari
>

please test the cairo packages from hardy-proposed

 status incomplete

 - Alexander

[Alexander Sack (asac)]

On Thu, Jun 19, 2008 at 11:55:03PM -0000, Kari Aliranta wrote:
>
> Well, I was wild and installed Nouveau driver for my Nvidia card to get EXA working. The result: about 15% improvement in the benchmark for FF3, but I get hugely improved (>2X) results in Gtkperf text scrolling benchmark and everything else is definitely faster, too. Very nice, but I wouldn't consider this a solution for the FF problem, though.
>

please test the cairo package from hardy-proposed

 affects ubuntu/xulrunner-1.9
 status incomplete

 - Alexander

[Kari Aliranta (kpaliran)]

> please test the cairo package from hardy-proposed

Tested with the nv driver, and doesn't help - cannot test with nouveau, as my X refuses to start with the latest snapshot.

Kari

[Kari Aliranta (kpaliran)]

Also, I got a chance to test this with the mga driver, and the problem is still there (also, the computer was a P3, and scrolling even plain text was definitely choppy).

Kari

[krom (krom)]

For me problem disappeared after installing latest nvidia beta drivers (180.06)

[Alexander Sack (asac)]

On Sun, Nov 16, 2008 at 04:03:21PM -0000, krom wrote:
> For me problem disappeared after installing latest nvidia beta drivers
> (180.06)
>

 affects ubuntu/xulrunner-1.9
 status invalid
 affects ubuntu/firefox
 status invalid
 affects ubuntu/linux
 status new

thanks

 - Alexander

[goto (gotolaunchpad)]

Why is this bug filed against linux? Firefox uses the whole CPU.

I think the JavaScript is very bad written or firefox has a bug or (what I assume most) there is a problem with the rendering. If the 'newsticker' moves, the whole thing must be new rendered. And if the graphics card can't do this, because the driver doesn't support this, then the CPU makes this and uses much CPU for this. The following comment supports this theory:

On Sun, Nov 16, 2008 at 04:03:21PM -0000, krom wrote:
> For me problem disappeared after installing latest nvidia beta drivers
> (180.06)

Really interesting! This shows, that this bug seems to be a bug in the graphics card's driver.

Maybe similar to bug 187383? (both try to render with cairo?)

[Robert (robrwo)]

How is this invalid? It's still a major problem for me.

On Sun, Dec 7, 2008 at 12:46 AM, Alexander Sack <email address hidden> wrote:
> ** Changed in: xulrunner-1.9 (Ubuntu)
> Status: Incomplete => Invalid
>
> ** Changed in: firefox (Ubuntu)
> Status: Incomplete => Invalid
>
> ** Also affects: linux (Ubuntu)
> Importance: Undecided
> Status: New
>
> --
> high cpu load because of JavaScript
> https://bugs.launchpad.net/bugs/185469
> You received this bug notification because you are a direct subscriber
> of the bug.
>

[goto (gotolaunchpad)]

I reopen the bug for firefox, because other browsers don't have this bug, so firefox causes this. If anyone knows better, please correct.

Would it make sense to open an Xorg bug for this?

[Michael Lustfield (michaellustfield)]

Marking invalid for Linux. This is not an issue with Linux.

[Michael Lustfield (michaellustfield)]

Can somebody give some evidence this is not an issues w/ the website rather than poor javascript/java code?

[Will Hui (willh126)]

This does not seem to be a problem for IE7 or Google Chrome on Windows XP. Both use 10% CPU or less when the site is loaded.

[Alexander Sack (asac)]

how does ffox on windows perform here?

[Alexander Sack (asac)]

i retitled this bug because generic titles just attract false-me-toos.

[Alexander Sack (asac)]

huge improvements have been made to javascript performance (including JIT) on 1.9.1 branch and trunk. Please check out the firefox-3.1 or even firefox-3.2 packages in fta's PPA and confirm whether its fixed there. Thanks ... https://edge.launchpad.net/~fta/+archive

[Michael Lustfield (michaellustfield)]

No response, assuming fixed for user.

[jtchil0 (jtchil0)]

I think I am seeing the same problem when the huffingtonpost.com website trying to update it's story. The computer comes to a halt for 10-30 seconds until the page has finished updating.