527728 - [meta] Track and improve scrolling latency

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Jeff Muizelaar [:jrmuizel]]

I have some tests of my early work here:
http://people.mozilla.com/~jmuizelaar/scroll/

Comment 1

[Jeff Muizelaar [:jrmuizel]]

If anyone has pages that scroll slowly please post urls here.

Comment 2

[Markus Stange [:mstange]]

data:text/html,<html%20style="-moz-box-shadow:%200%200%20150px%20black%20inset;%20height:%205000px"> ;)

Comment 3

[(no longer active)]

Is this Mac specific?

Comment 4

[(no longer active)]

One instance that I've noticed on Linux is scrolling the main gmail conversation view when an instance of the in-window chat session is open.

The chat session is (I think) basically a div with position absolute, high z-index, and bottom and right set to 0.

Comment 5

[Jeff Muizelaar [:jrmuizel]]

(In reply to comment #4)
> One instance that I've noticed on Linux is scrolling the main gmail
> conversation view when an instance of the in-window chat session is open.

This is bug 528604.

Comment 6

[Drew Willcoxon :adw]

Slow to scroll :(
http://design-challenge.mozillalabs.com/jetpack-for-learning/

Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.2b3pre) Gecko/20091113 Namoroka/3.6b3pre

Comment 7

[Karl Tomlinson (:karlt)]

(In reply to comment #6)
> http://design-challenge.mozillalabs.com/jetpack-for-learning/

-moz-box-shadow: 0pt -3px 0px rgba(200, 200, 200, 0.3) inset, 0pt -3px 6px rgb(255, 255, 255) inset, 0pt 20px 40px rgb(255, 255, 255) inset, 0pt 2px 8px rgba(200, 200, 200, 0.3);

Comment 8

[(no longer active)]

I filed another scrolling perf problem as bug 529096.

Comment 9

[WL]

Google reader with extra long paragraph or a lot of images will make the scrolling slow.
My platform:Win32,3.7nightly @ 1920*1080 resolution

Comment 10

[Peter Retzer (:pretzer)]

I have a account for the "google wave preview" and scrolling seems to be very slow in the Wave-Content-Area for me. 
But that might be connected to the "different" scrolling-mechanism they implemted there.

Comment 11

[Rob Campbell [:rc] (:robcee)]

Feedly. Very slow for me in Win7 with smooth scrolling enabled.

Comment 12

[(dormant account)]

gmail scrolling is painful when a chat box is opened

Comment 13

[(no longer active)]

(In reply to comment #12)
> gmail scrolling is painful when a chat box is opened

Yes, we filed bug 528604 on it, and it has a reviewed patch! :-)

Comment 14

[Chris Smith]

we did some tests with scrolling for different OSs some time back:

http://grigio.org/scroll_benchmark_really_firefox_linux_so_bad

also my own benchmark, similar to "~jmuizelaar"

http://sliste.sl.funpic.de/bmark/

Hope this helps!

Comment 15

[Andrew Stibbard]

Not sure if it's what you're after, but this page exhibits a lot of vertical tearing and stutter when scrolling down:

 http://nwn2db.com/builder.php

Other browsers (IE8, Safari 4, Chrome 2+) scroll smoothly. Other areas of the site are stuttery too (eg. 'Library' menu on the front page) so I'm tempted to call this a redraw/dirty rect problem.

Comment 16

[Jeff Muizelaar [:jrmuizel]]

(In reply to comment #9)
> Google reader with extra long paragraph or a lot of images will make the
> scrolling slow.
> My platform:Win32,3.7nightly @ 1920*1080 resolution

This seems to be related to box shadow and is a regression from 1.9.2

Comment 17

[Jeff Muizelaar [:jrmuizel]]

(In reply to comment #15)
> Not sure if it's what you're after, but this page exhibits a lot of vertical
> tearing and stutter when scrolling down:
> 
>  http://nwn2db.com/builder.php
> 
> Other browsers (IE8, Safari 4, Chrome 2+) scroll smoothly. Other areas of the
> site are stuttery too (eg. 'Library' menu on the front page) so I'm tempted to
> call this a redraw/dirty rect problem.

Looks like this could be box shadow related too.

Comment 18

[Andrew Stibbard]

(In reply to comment #17)
> Looks like this could be box shadow related too.

Nah, only one reference to -moz-box-shadow in the CSS, and we broke it intentionally after we suspected it to be the performance problem.

There's a dramatic improvement after removing the fixed-pos black bar at the bottom (#utilBar), or a slight improvement after just removing the fixed-pos alpha png gradient (#utilFade).

http://xhva.net/work/css3/  has similiar perf problems with fixed-pos alpha pngs (#header and #footer) but admittedly it uses -moz-box-shadow and CSS columns which may complicate it.

Comment 19

[Karl Tomlinson (:karlt)]

(In reply to comment #15)
>  http://nwn2db.com/builder.php

There are big improvements (or changes at least) between 1.9.1 and 1.9.2 here.
What OS and Mozilla Browser are you using.

Comment 20

[(no longer active)]

(In reply to comment #18)
> http://xhva.net/work/css3/  has similiar perf problems with fixed-pos alpha
> pngs (#header and #footer) but admittedly it uses -moz-box-shadow and CSS
> columns which may complicate it.

If this matters, I noticed that the performance on this page is much worse when fully zoomed in.

Comment 21

[(no longer active)]

(In reply to comment #19)
> (In reply to comment #15)
> >  http://nwn2db.com/builder.php
> 
> There are big improvements (or changes at least) between 1.9.1 and 1.9.2 here.
> What OS and Mozilla Browser are you using.

I think the performance of scrolling in this page is better with the fix to bug 528604, since this page also uses a div with a fixed position.

Comment 22

[Andrew Stibbard]

(In reply to comment #19)
> There are big improvements (or changes at least) between 1.9.1 and 1.9.2 here.
> What OS and Mozilla Browser are you using.

XP 32bit, tested in both Firefox 3.5.5 and the 20091111 nightly. There's a definite improvement in the nightly, yes, but it's still slower than competitors.

Comment 23

[Jeff Muizelaar [:jrmuizel]]

(In reply to comment #22)
> (In reply to comment #19)
> > There are big improvements (or changes at least) between 1.9.1 and 1.9.2 here.
> > What OS and Mozilla Browser are you using.
> 
> XP 32bit, tested in both Firefox 3.5.5 and the 20091111 nightly. There's a
> definite improvement in the nightly, yes, but it's still slower than
> competitors.

Andrew,

If you could come up with a minimal test page that shows the slowness difference between different browsers that would be tremendously helpful. Further, we could  add it to a test suite to ensure that we don't regress it.

Comment 24

[Jeff Muizelaar [:jrmuizel]]

(In reply to comment #9)
> Google reader with extra long paragraph or a lot of images will make the
> scrolling slow.
> My platform:Win32,3.7nightly @ 1920*1080 resolution

The google reader problem is bug 530686

Comment 25

[Paul Stone]

Scrolling in this SVG / SMIL demo is pretty bad. Click one of the icons at the bottom and then scroll:

http://people.mozilla.org/~dholbert/demos/smildock/01/smildock.svg

Comment 26

[Jeff Muizelaar [:jrmuizel]]

(In reply to comment #25)
> Scrolling in this SVG / SMIL demo is pretty bad. Click one of the icons at the
> bottom and then scroll:
> 
> http://people.mozilla.org/~dholbert/demos/smildock/01/smildock.svg

Bug 534721.

Comment 27

[Paul Stone]

The new version of Twitter (currently in preview) scrolls pretty badly on the latest trunk nightly. The new layout contains several position:fixed elements with box-shadow. The slowness isn't all the fault of box-shadow though - disabling those CSS rules with Firebug speeds up scrolling a little, but it's still slow.

Comment 28

[Paul Stone]

Sorry, should have mentioned - that's on WinXP with no hardware acceleration.

Comment 29

[Theodore]

In Google Reader, click the 'Explore' button on the left, notice that smooth scrolling is unbearably slow. I'm running Firefox 4 RC on Windows 7 x64 with HW acceleration on.

Comment 30

[Stuart Cook]

Twitter is still very slow, at least on OS X.

http://www.giantbomb.com and http://www.tested.com are also quite slow, especially near the top of the page.

Comment 31

[Benoit Girard (:BenWa)]

I think a lot of these bugs can be closed since they are solved by Layers. The only problem I noticed are with mouse overs and plug-in playing the performance is still sub-optimal.

Comment 32

[CruNcher]

Addon Manager seems also affected

Comment 34

[Gregory Pappas [:gregp]]

Inactive metabug