1169016 - Enabling video controls decreases app performace

Status: RESOLVED WORKSFORME

(Core :: Audio/Video: Playback, defect, P3)

Description

[Nicholas Troast [:ntroast]]

When video controls are enabled in a packaged application I see that the app uses 5x more CPU time even if the controls are hidden. I used kernel trace logs to collect the CPU time data.

STR:
Create a simple packaged app with the following body:

<video controls autoplay src="http://html5video.org/players/media/sintel.3gp"></video>

Remove the controls attribute and the app consumes less CPU time

Gecko: 197cb3859f55ca587256c76648adec54a1f502ef
Gaia: 732acec6f37d13ccea6b0ddc48904a53a2970894

Comment 1

[Bobby Chien]

steven, could you help to investigate this?

Comment 2

[StevenLee[:slee]]

(In reply to Nicholas Troast [:ntroast] from comment #0)
> When video controls are enabled in a packaged application I see that the app
> uses 5x more CPU time even if the controls are hidden. I used kernel trace
> logs to collect the CPU time data.
> 
> STR:
> Create a simple packaged app with the following body:
> 
> <video controls autoplay
> src="http://html5video.org/players/media/sintel.3gp"></video>
> 
> Remove the controls attribute and the app consumes less CPU time
Hi Nichola,
Can you provide the statistics data? I tried on my desktop. When I enable "controls", most of the CPU time spends on garbage collection threads. I want to make sure we are looking into the same thing. Thanks.

Comment 3

[Nicholas Troast [:ntroast]]

Hi Steven,

I would share the kernel log data, but I am not sure that it would be very informative. The only process that is consuming more cpu is the app itself. All other processes look normal.

Comment 4

[StevenLee[:slee]]

(In reply to Nicholas Troast [:ntroast] from comment #3)
> I would share the kernel log data, but I am not sure that it would be very
> informative. The only process that is consuming more cpu is the app itself.
> All other processes look normal.
I think you are using ftrace to get the data, right? The output of freace is calculated per thread. I want to make sure FxOS and desktop browser are the same. Thanks.

Comment 5

[Jayme Mercado [:JMercado] (Inactive 3/4/2016)]

Bobby what are you looking to have done for the qawanted request, branch checks?  If so can you provide some test apps (one with and one without video controls) for comparison purposes.

Comment 6

[Bobby Chien]

Nicholas, could you provide Jayme's request in comment 5? So we could have test to against branches. Thanks.

Comment 7

[howie [:howie]]

Hi Nicholas, can we have your feedback here? Thank you

Comment 8

[Nicholas Troast [:ntroast]]

(In reply to Bobby Chien [:bchien] from comment #6)
> Nicholas, could you provide Jayme's request in comment 5? So we could have
> test to against branches. Thanks.

I provided the test app in the bug description. It is a single line.

Comment 9

[Hermes Cheng (inactive after July 27, 2015)]

Are we going to fix this for v2.2?

Comment 10

[Josh Cheng [:josh]]

[Blocking Requested - why for this release]:
2.2 CCed. Continue investigate in next release.

Comment 11

[Al Tsai [:atsai]]

[Triage] Minus it since it's partner related. We'll rely on partner's help.

Comment 12

[Anthony Jones (:ajones, :kentuckyfriedtakahe, :k17e)]

Mass change P2 -> P3

Comment 13

[Alastor Wu [:alwu]]

Considering this is an pretty old bug, and we've changed video control a lot, and I didn't see this issue on the latest Release. So close this bug.