769003 - SVG image causes very high lag when switching tabs/minimizing

Status: REOPENED

(Core :: Graphics, defect)

Description

[Josh Matthews [:jdm]]

nemo points out that visiting the given SVG image causes a very high lag in nightlies when switching tabs and minimizing the browser. This lag is not present in 3.6, but the regression window could use a bit more work.

Comment 1

[Daniel Holbert [:dholbert]]

WFM, using 64-bit Linux nightly. (no noticeable lag when switching tabs, or when minimizing)

  Mozilla/5.0 (X11; Linux x86_64; rv:16.0) Gecko/16.0 Firefox/16.0
  Built from http://hg.mozilla.org/mozilla-central/rev/5cdbeae14405

Comment 2

[nemo]

Heavy lag for me, still, in a clean profile, latest Firefox Linux 64 bit nightly and Firefox Windows 32 bit Nightly.

On the Linux machine, which is considerably more powerful, switching away from Firefox to this window to type this up, has about a two second lag focusing/unfocusing the window as long as the SVG tab is active.  Switching to the tab from another open tab is similarly slow.

On the Windows machine, which is not as powerful, the lag is more like 5 seconds.

I suppose you not noticing anything where we do could be due to some difference in rendering config, but maybe you just have a really really powerful machine :)

I pulled out one of our test Android devices, a Samsung Note and fired up Firefox Stable.

Switching to the tab with the SVG took 5-7 seconds.
Switching away from it was ridiculously slow the first time (like, 15+ seconds) but after that seems to have been fast, so, not sure what was going on there.

I have noticed that the slowness can happen both on switching away and switching to, though.

Comment 3

[nemo]

Slightly smaller regression window.  Grabbing the latest nightly from the last time I did bisecting on this XP machine, Firefox 4.0b from 2010-09-08 loads instantly with no delays.

Comment 4

[Daniel Holbert [:dholbert]]

Still no noticable lag for me, latest Firefox Linux 64 bit nightly on a Dell Precision 3.50GHz w/ Ubuntu 12.04.  (powerful, but not ridiculously so)

I tried Firefox release & nightly on my Android Nexus S, too, and I get 1-2s delay on rendering the content when switching _to_ the tab (which is reasonable/expected, I think), and no UI lag or delay on switching away from the tab.

So, still WORKSFORME. :)  I'll can give it a try on my (less powerful) mac mini next time I'm in Mountain View.

Comment 5

[nemo]

Ok. This is really strange.
If I try my win32 nightly that I saved on this machine from 2010-09-08, I can flip back between tabs instantly, no noticeable delay (contrated w/ more like 5 seconds in the latest nightlies on that same machine)

But, on my more powerful x86_64 linux machine, the delays (which are more like a bit under two seconds, and involve noticeable missing chunks of unpainted window as I switch back and forth from the firefox window to other apps) are visible in 2010-01-01 and not in 2008-06-01 (32 bit - no 64 bit builds back then).

I'm kinda stuck at narrowing that down, since even 2008-09-01 aborts due to missing libdbus-glib-1.so.2 in 32bit version.

Soooo. Below, where fast is basically "renders instantly"
Windows XP 32bit         - 2010-09-08 fast  2012-06-01 slow  (5 seconds to switch to the tab for slow)
Ubuntu Linux 12.04 64bit - 2008-06-01 fast  2010-01-01 slow  (1-2 seconds to switch to the tab for slow, focusing another app causes noticeable ugly paint delays in the second app)

Kinda confused by this.

And, 1-2 seconds lag doesn't sound that great really.
If I take my Samsung Galaxy Note where switching to the tab takes 5 seconds in Firefox Stable, and open, oh:
http://www.w3.org/Graphics/SVG/Test/20080912/svggen/jpeg-required-201-t.svg  (an SVG with an embedded jpeg)

I can switch to and from that tab pretty much instantly. No noticeable lag.

And, FWIW, Chromium stable and Opera stable both load and switch to-from the tab instantly with no delay at all on both the XP and Linux machines.  So if you don't really believe this is a regression you could at least accept it as parity :-p

BTW.  This SVG used to be an example of some sucky SVG rendering in Chromium, although they do seem to have improved it a bit (the stripes render at least, even if they are not as attractively blended as in Firefox and Opera)

Comment 6

[(mostly gone) XtC4UaLL [:xtc4uall]]

I can confirm Performance Issues described as above with HWA disabled against recent Nightlys.

Using HWA (i.e. D2D/D3D10, Ati 4580, Win 7 x64) this is a non-Issue.

Testing against Firefox 4 OTH there's no Issue no Matter of which HWA Settings.

Comment 7

[(mostly gone) XtC4UaLL [:xtc4uall]]

(In reply to XtC4UaLL [:xtc4uall] from comment #6)
Central:
Last good nightly: 2012-04-20
First bad nightly: 2012-04-21

Pushlog: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=22bfdebf5cae&tochange=7fda4d02d3ca

Inbound:
Last good nightly: 2012-04-19
First bad nightly: 2012-04-20

Pushlog: http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=e9e62564ab34&tochange=df08546b7cc4

nemo, can you please check if a) your Issue is with HWA disabled and b) the Ranges match your Issue.
Else I guess I should file a new Report.

Comment 8

[Alice0775 White]

Regression window(m-i) with HWA disabled
Good:
http://hg.mozilla.org/integration/mozilla-inbound/rev/7b59d08bd69a
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:14.0) Gecko/20120419 Firefox/14.0a1 ID:20120419203018
Bad:
http://hg.mozilla.org/integration/mozilla-inbound/rev/dcbe20a6f53a
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:14.0) Gecko/20120419 Firefox/14.0a1 ID:20120419211922
Pushlog
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=7b59d08bd69a&tochange=dcbe20a6f53a

Triggered by:
dcbe20a6f53a	Jeff Muizelaar — Bug 746695. Update libpixman to 0.24.4 a=joe

Comment 9

[nemo]

m'k.
Well, I suppose now that it is identified, not much help, but...
On my Android Acer A500 tablet running Ice Cream Sandwich:
Firefox 10: 2-3 seconds to switch to svg tab
Nightly:    10 seconds to switch to svg tab

I tried "Check for Updates" but Firefox stable said there weren't any *shrug*  guess 10 is latest for this hardware/os?

Comment 10

[nemo]

Triggered by:
dcbe20a6f53a    Jeff Muizelaar — Bug 746695. Update libpixman to 0.24.4 a=joe

Does this mean this bug should be reassigned to him to look at?

Comment 11

[Ondřej Hruška]

I know this bug is slow, nevertheless the lag when switching to and fro a SVG tab is still present in version 20. 
[Linux]

Comment 12

[Ondřej Hruška]

I know this bug is old, nevertheless the lag when switching to and fro a SVG tab is still present in version 20. 
[Linux]

Comment 13

[Justin [:JW_SoftvisionQA]]

I can not reproduce this issue on  	Mozilla/5.0 (Windows NT 6.3; WOW64; rv:48.0) Gecko/20100101 Firefox/48.0

I will close this issue as wfm. Please reopen if you can still reproduce.

Comment 14

[(mostly gone) XtC4UaLL [:xtc4uall]]

Yep, as per Comment 6 still a noticable 1-2s lag with tab switching in the HWA-less/safemode case testing Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:48.0) Gecko/20100101 Firefox/48.0 ID:20160407030234 CSet: b6683e141c47c022598c0caac3ea8ba8c6236d42

Adapter Description: ATI Radeon HD 4800 Series
Adapter Drivers: aticfx64 aticfx64 aticfx64 aticfx32 aticfx32 aticfx32 atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64
Adapter RAM: 512
Asynchronous Pan/Zoom: wheel input enabled; touch input enabled
ClearType Parameters: D [ Gamma: 1800 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 400 ] D [ Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 300 ]
Device ID: 0x9442
DirectWrite Enabled: false (10.0.14316.1000)
Driver Date: 1-13-2015
Driver Version: 8.970.100.9001
GPU #2 Active: false
GPU Accelerated Windows: 0/1 Basic (OMTC)
Subsys ID: 026a1043
Supports Hardware H264 Decoding: No; Hardware video decoding disabled or blacklisted
Vendor ID: 0x1002
WebGL Renderer: Google Inc. -- ANGLE (ATI Radeon HD 4800 Series Direct3D11 vs_4_1 ps_4_1)
windowLayerManagerRemote: true
AzureCanvasAccelerated: 0
AzureCanvasBackend: skia
AzureContentBackend: cairo
AzureFallbackCanvasBackend: cairo

Comment 15

[Gregory Pappas [:gregp]]

Nemo, are you still experiencing this issue?

Comment 16

[nemo]

I'll try a test in next day or two. Man this is an old bug. Checking the comments it appears it's about disabling hardware acceleration and using software-only, which has caused some spurious "WFM". So, I guess I need to retest in clean profile with all graphics acceleration disabled.