1312890 - OpenFL applications run slowly in Firefox 49.0.2 on Mac

Status: RESOLVED WORKSFORME

(Core :: General, defect)

Description

[curtis.h]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_9_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/52.0.2743.116 Safari/537.36

Steps to reproduce:

1.  Update Firefox to version 49.0.2
2.  Start Firefox by clicking on the desktop icon.
3.  (Optionally) Open Activity Monitor from Finder
4.  Paste http://demo-north-america.s3.amazonaws.com/mirrormagic/m4/html5/index.html in the Firefox address bar and press Enter.
5.  Wait for game to finish loading.
6.  Press the skip button (->) until the intro sequence is finished and the game symbols appear.
6.  Click on the demo button (top left corner).
7.  Click Pic4 in the demo controls panel.


Actual results:

Game runs somewhere between 10-40 FPS on Mac, and increases CPU usage over 100%.  Using this version of Firefox on PC runs ~60 FPS.  Using the previous version (49.0.1) on Mac also runs ~60 FPS.

Performance Profile:
https://cleopatra.io/#report=fa53f62a5a52f5ae2bea4fe3c120cf0158a95e69


Expected results:

Game should run at ~60 FPS.

Comment 1

[Loic]

Could you find a regression range by using the tool mozregression maybe.
http://mozilla.github.io/mozregression/

Comment 2

[curtis.h]

Here are the results from mozregression:

INFO: Looks like the following bug has the changes which introduced the regression:
https://bugzilla.mozilla.org/show_bug.cgi?id=1272890

Comment 3

[Loic]

(In reply to curtis.h from comment #2)
> Here are the results from mozregression:
> 
> INFO: Looks like the following bug has the changes which introduced the
> regression:
> https://bugzilla.mozilla.org/show_bug.cgi?id=1272890

This regression range is probably wrong. Do you remember if the performance when rendering the game was better in previous versions of Firefox?

If yes, rerun mozregression with an older good release, like --good=46.

Comment 4

[Adrian Florinescu [:aflorinescu]]

I agree with Loic that the regression range seems wrong, that change (Bug 1272890) was even backed out from Nightly52, so I don't see any way in which it would've made to 49.0.2.

I've ran a few tests on the demo provided and 49, 49.0.1 and 49.0.2 have the same behavior. It starts around 52fps and as you let it run, it starts dropping constantly and the performance of the browser goes down the drain as time passes.

Comment 5

[curtis.h]

I tried rerunning the regression with a larger window, and got the following:

INFO: Looks like the following bug has the changes which introduced the regression:
https://bugzilla.mozilla.org/show_bug.cgi?id=1294812

Comment 6

[Loic]

(In reply to curtis.h from comment #5)
> I tried rerunning the regression with a larger window, and got the following:
> 
> INFO: Looks like the following bug has the changes which introduced the
> regression:
> https://bugzilla.mozilla.org/show_bug.cgi?id=1294812

This bug landed only in FF52, so it can't be the possible source of a regression in FF49.

Comment 7

[curtis.h]

Regardless, those were the results that were produced by mozregression.

Having downloaded and tested the issue on the 50.0b11 build, I could not reproduce it, so this may all be moot in a couple weeks.

Comment 8

[Adrian Florinescu [:aflorinescu]]

I agree with Loic on Comment6, that can't be possibly the culpable for this issue. Let's review this issue once 50 reaches release. 
Curtis, could try in two weeks time a reproducibility status for this bug on all channels?

Comment 9

[curtis.h]

Will do.

Comment 10

[curtis.h]

We've received new data from our client that the tests we were performing do not replicate the issue, and that the problem may have existed prior to the 49.0.2 update.  Please close this ticket.

Comment 11

[Firefox Bug Husbandry Bot]

Moving from Core::Untriaged to Core::General https://bugzilla.mozilla.org/show_bug.cgi?id=1407598