Closed Bug 891916 Opened 11 years ago Closed 9 years ago

Vimeo videos image is delayed

Categories

(Core :: Graphics, defect)

Product:
Core
Component:
Graphics
Platform:
x86_64
Windows 7
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED FIXED
Tracking Flags:
Tracking Status
firefox23 - affected
firefox24 - affected
firefox25 - affected

People

(Reporter: sbadau, Assigned: bas.schouten)

References

Details

Mozilla/5.0 (Windows NT 6.1; WOW64; rv:25.0) Gecko/20130709 Firefox/25.0
Build ID: 20130709030204

1. Navigate to https://vimeo.com/
2. Play any of the videos

Expected results:
The video is playing properly.

Actual results:
The image is delayed by compared to the sound and is jerky.
If I Copy the video URL and I play the video from the copied link, the video is playing just fine (image and sound are correlated).
Please see the screencast to see the jerky image.
http://screencast.com/t/5nrbu8xUXTJ

Notes:
1. Reproducible on the latest Nightly 25.0a1, on the latest Aurora 24.0a2, on Firefox 23 beta 4 and ever since Firefox 9.0.1 (the first version of Firefox where Vimeo videos can be played).
2. I could only reproduce the issue on a PC running with an ATI Radeon HD 5450 video card (the same issue is not reproducible on Opera 15.0.1147.141 and on Chrome 27.0.1453.116 m). 
3. I could not reproduce the issue on a PC running with AMD Radeon HD 7700 Series video card or nVIDIA GeForce 8400 video card.
4. Reproduced with the following Flash versions:
11.8.800.94, 11.7.700.232, 11.6.602.180, and 11.8.800.95 beta.
(In reply to Simona B [QA] from comment #0)
> 2. I could only reproduce the issue on a PC running with an ATI Radeon HD
> 5450 video card (the same issue is not reproducible on Opera 15.0.1147.141
> and on Chrome 27.0.1453.116 m). 

If those machines are otherwise similar, what about disabling hardware in Firefox and/or Flash on the affected one?
(In reply to Georg Fritzsche [:gfritzsche] from comment #1)
> (In reply to Simona B [QA] from comment #0)
> > 2. I could only reproduce the issue on a PC running with an ATI Radeon HD
> > 5450 video card (the same issue is not reproducible on Opera 15.0.1147.141
> > and on Chrome 27.0.1453.116 m). 
> 
> If those machines are otherwise similar, what about disabling hardware in
> Firefox and/or Flash on the affected one?

The issue is still reproducible after disabling the hardware acceleration in Flash but disabling the hardware acceleration in Firefox does the trick and the issue is not reproducible anymore.
BenWa, i think you have some experience with gfx in that context?
Do you have an idea on what we could do here?
Flags: needinfo?(bgirard)
Looks like a performance issues. Can we get a performance profile?

https://developer.mozilla.org/en-US/docs/Performance/Reporting_a_Performance_Problem
Flags: needinfo?(bgirard) → needinfo?(simona.marcu)
(In reply to Benoit Girard (:BenWa) from comment #4)
> Looks like a performance issues. Can we get a performance profile?
> 
> https://developer.mozilla.org/en-US/docs/Performance/
> Reporting_a_Performance_Problem

http://people.mozilla.com/~bgirard/cleopatra/#report=e4a484985c901e86b6a9ba54ce8537cf8dc02283
Flags: needinfo?(simona.marcu)
Can you turn off d2d from the preferences, restart the browser and try again? If it's still not smooth please link a new profile.
(In reply to Benoit Girard (:BenWa) from comment #6)
> Can you turn off d2d from the preferences, restart the browser and try
> again? If it's still not smooth please link a new profile.

After setting gfx.direct2d.disabled to true - the video plays fine.
CCing :bas for d2d bug.
This looks to be caused by the two particle emitters on the site. (One at the top and one at the bottom). They are drawing a bunch of particles with OPERATOR_ADD which makes D2D sad.
While not strictly a regression this seems like something we may want to track.
status-firefox23: --- → affected
status-firefox24: --- → affected
status-firefox25: --- → affected
tracking-firefox23: --- → ?
tracking-firefox24: --- → ?
tracking-firefox25: --- → ?
(In reply to Anthony Hughes, Mozilla QA (:ashughes) from comment #10)
> While not strictly a regression this seems like something we may want to
> track.

It's a regression from turning on d2d.
(In reply to Anthony Hughes, Mozilla QA (:ashughes) from comment #10)
> While not strictly a regression this seems like something we may want to
> track.

We'd consider tracking if we had more information about which of the popular graphics cards are affected - apparently ATI Radeon HD 5450, but can we get more info about the most popular cards by users?

Daniel do you have any metrics data that could help here?
Assignee: nobody → bas
Flags: needinfo?(deinspanjer)
Keywords: qawanted
Simona, while we wait to hear back from Daniel do you have any other hardware in-house that you could test this with? It would be good to do some cataloging of this bug across available hardware.
QA Contact: simona.marcu
I could not reproduce the issue with the following video cards:
- AMD Radeon HD 6450
- Intel HD 2500 
- nVIDIA GeForce GT 610
- nVIDIA GeForce 8400
- AMD Radeon HD 7700 Series
(In reply to Simona B [QA] from comment #14)
> I could not reproduce the issue with the following video cards:
> - AMD Radeon HD 6450
> - Intel HD 2500 
> - nVIDIA GeForce GT 610
> - nVIDIA GeForce 8400
> - AMD Radeon HD 7700 Series

I'm very surprised by this. This is general D2D problem. It might be worse on some hardware but the problem should be completely hardware independent.

I'm also don't really think there's much to track here either. This performance problem has existed since Firefox 4 and the reason it showed up is because vimeo added those canvas particle emitters recently. The most valuable thing to do would probably be to reach out to vimeo and ask them to stop.
Agreed, let's try to reach out to Vimeo ni? on Kev, Jason, and Lawrence to see if someone has a contact for them.
Flags: needinfo?(lmandel)
Flags: needinfo?(kev)
Flags: needinfo?(jsmith)
(In reply to lsblakk@mozilla.com [:lsblakk] from comment #16)
> Agreed, let's try to reach out to Vimeo ni? on Kev, Jason, and Lawrence to
> see if someone has a contact for them.

I'm going to bounce my needinfo request over to Karl, as he's been doing a lot of outbound communication now for the web compatibility team.

Karl - Can you outreach to Vimeo about this bug?
Flags: needinfo?(jsmith)
Flags: needinfo?(kdubost)
Clearing previous ni? Christopher Arnold stepped up to reach out.
Flags: needinfo?(lmandel)
Flags: needinfo?(kev)
Flags: needinfo?(deinspanjer)
Flags: needinfo?(carnold)
Flags: needinfo?(kdubost)
We should have someone on their team to look at this soon.
Flags: needinfo?(carnold)
I have read all the comments.
Before reaching out vimeo, it would be good to know which piece of code you are referring to and how do you think it could be fixed. It's better to have a practical suggestion. :) It helps specifically when the issue is not visible in some other browsers.
Removing qawanted until comment 16, 20 get accomplished.
Keywords: qawanted
(In reply to Karl Dubost from comment #20)
> I have read all the comments.
> Before reaching out vimeo, it would be good to know which piece of code you
> are referring to and how do you think it could be fixed. It's better to have
> a practical suggestion. :) It helps specifically when the issue is not
> visible in some other browsers.

On the front page of vimeo.com they have these little particle emitters on the doors with 'V's. Those cause quite a bit of slow down in Firefox on Windows with D2D because of globalCompositeOperation = 'add'
Ok understand the issue about the V which seems to slow down *only certain* versions of Firefox. What should be changed on vimeo code for:

1. maintaining the particle emitters
2. and solving the performance issue on Firefox.

It will help a lot contacting them with a patch. :)
Thanks.
Flags: needinfo?(jmuizelaar)
(In reply to Karl Dubost from comment #23)
> Ok understand the issue about the V which seems to slow down *only certain*
> versions of Firefox. 

All recent versions of Firefox are affected, but only those using the D2D backend on Windows.

There's not really a great way to maintain the current effect. The suggestion is really just to remove it because of the negative performance impact it has on Firefox windows users.

I'll try to look at IE9 tomorrow to see if it has a similar problem.
Flags: needinfo?(jmuizelaar)
Vimeo team are aware of this issue.  Their response was:

We're in the process of rolling out a new version of our player that adds support for HTML5 in Firefox where H.264 is supported. I would hope that that will help the issue as well.
(In reply to Christopher Arnold from comment #25)
> Vimeo team are aware of this issue.  Their response was:
> 
> We're in the process of rolling out a new version of our player that adds
> support for HTML5 in Firefox where H.264 is supported. I would hope that
> that will help the issue as well.

This will likely not make any difference. The performance problem comes particle emitter animation above the V on the door and not video as far as I can tell.
I guess Core::Graphics is a better fit for this bug.
Component: Plug-ins → Graphics
We've refactored our particle emitter class so it should use less memory and also switched from "lighter" to "normal" as the composite operation. If you are still noticing performance issues on Win/FF, let us know and we will remove the effect for that specific group of users.
Depends on: 902952
This should be fixed now that we use D2D1.1
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.