1164648 - Web browsing on Firefox OS is not as efficient as Android

Status: RESOLVED WONTFIX

(Firefox OS Graveyard :: Performance, defect, P2)

Description

[Nicholas Troast [:ntroast]]

Attachment: browser_comparison.pdf

Gaia: 83a63e0e6fcc22c6a74b06ef77b88d5049719cad
Gecko: 4f049c08e74aab39ee81e7c6e7fcee311a1868b0

The BrowserMark process is a simple packaged web app that sets a 5 second interval to change the source of an iframe. A random number is used to prevent caching.

setInterval(function () {
  $('iframe').attr('src', 'http://www.wikipedia.org/wiki/Firefox?rand=' + Date.now());
}, 5000);

5x 60second kernel traces were averaged and divided by 60 in order to produce the attached pdf. So each value is the time in ms out of one second.

The expectation is that Firefox OS uses equal or less CPU time than Android while browsing.

Comment 1

[Dylan Oliver [:doliver]]

Thinker, can you take a look at this?

Comment 2

[Thinker Li [:sinker]]

I don't really understand the numbers.  Are they the process's CPU time in ms?

Comment 3

[Nicholas Troast [:ntroast]]

Yes the numbers are the process's CPU time in ms over a period of one second.

Comment 4

[Ken Chang[:kenkai]]

(In reply to Nicholas Troast [:ntroast] from comment #3)
> Yes the numbers are the process's CPU time in ms over a period of one second.

May I know how many CPUs are running when you run this app and what are the frequencies of CPUs?

Comment 5

[Nicholas Troast [:ntroast]]

(In reply to Ken Chang[:ken] from comment #4)
> May I know how many CPUs are running when you run this app and what are the
> frequencies of CPUs?

4 cores @ 1.3GHz

Comment 6

[Dylan Oliver [:doliver]]

Ken, is your team looking at this one?

Comment 7

[Ken Chang[:kenkai]]

(In reply to Dylan Oliver [:doliver] from comment #6)
> Ken, is your team looking at this one?
Hi Dylan,
We aren't confident that we can fix this issue in 2.2. But Thinker's team will keep looking into this problem.

Comment 8

[Bobby Chien]

[Tracking Requested - why for this release]:

Noted as performance wanted bug to be care in future release. I suggest not in v2.2.

Comment 9

[Ken Chang[:kenkai]]

(In reply to Nicholas Troast [:ntroast] from comment #5)
> (In reply to Ken Chang[:ken] from comment #4)
> > May I know how many CPUs are running when you run this app and what are the
> > frequencies of CPUs?
> 
> 4 cores @ 1.3GHz

Hmm, that means that Firefox is run by 4 cores with 1.3 GHz while doing this test.
And another thing I learned from Dylan is that this problem is relevant to the power
consumption. I wonder if we could have your power consumption datums of Android and Firefox
while doing this test.

Comment 10

[Bobby Chien]

Ni Dylan and Nicolas, should we move it out from v2.2?

Comment 11

[Dylan Oliver [:doliver]]

Confirmed that we will move this off the 2.2 list. Bumping to 3.0? to retain visibility.

Comment 12

[Dylan Oliver [:doliver]]

Removing dependency on CAF 2.2, per agreement with partner.

Comment 13

[Dylan Oliver [:doliver]]

Hi Thinker,

Is there any progress on this issue? Or are there maybe some other more specific bugs that we should be watching for 2.5?

Comment 14

[Dylan Oliver [:doliver]]

Ken, see comment #13 -- is there some activity on this topic that we can/should be tracking for 2.5?

Comment 15

[Bobby Chien]

Dylan, we haven't start profile in bug yet. I will request issue profile. Thanks.

Comment 16

[Dylan Oliver [:doliver]]

Clearing blocking request for now and will leave it to Bobby to nominate specific work that needs to be included based on the analysis.