1043857 - Follow up to bug 1042083, launch time regression

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::System, defect)

Comment 1

[Etienne Segonzac (:etienne)]

Attachment: Gaia PR

Comment 2

[(inactive after 6/18) Alive Kuo [:alive]]

Comment on attachment 8462432 [details] [review]
Gaia PR

Delay the instantiation until app is opened sounds necessary.. cannot think of anything harmful.

Comment 3

[Etienne Segonzac (:etienne)]

https://github.com/mozilla-b2g/gaia/commit/8532cb66d11048c88f2fc5fc7bc72be708ae153c

Comment 4

[Miller Medeiros [:millermedeiros]]

I'm currently looking at a performance regression on the calendar app (Bug 1059349) and it looks like this bug is related to it.. even tho the performance improved, it is still not as fast as before Bug 1042083 landed (still ~100ms slower than before).

as you can see on datazilla[1]:

Gaia Rev: c419e71084ae5e4f | Avg:1067 | Med:1101 | Min:992  | Max:1116
Gaia Rev: d15e133c0bac9f22 | Avg:1247 | Med:1188 | Min:1161 | Max:1440
Gaia Rev: 8532cb66d11048c8 | Avg:1168 | Med:1155 | Min:1129 | Max:1228

 [1] https://datazilla.mozilla.org/b2g/?branch=master&device=flame-319MB&range=60&test=startup_%3E_moz-app-visually-complete&app_list=calendar&app=calendar&gaia_rev=8532cb66d11048c8&gecko_rev=beaa21982270&plot=avg

Comment 5

[Etienne Segonzac (:etienne)]

(In reply to Miller Medeiros [:millermedeiros] from comment #4)
> I'm currently looking at a performance regression on the calendar app (Bug
> 1059349) and it looks like this bug is related to it.. even tho the
> performance improved, it is still not as fast as before Bug 1042083 landed
> (still ~100ms slower than before).
> 

100ms seems high, but it's completely possible that the WindowManager work around this time had a launch time measurement impact. We're touching code running closely to the launch time events so it's always tricky.

That said, I don't think there's any other low-hanging fruits performance wise when it comes to the new appChrome/appTitleBar.

Basically we delay as much as we can until we get the "mozbrowserloadend" event _or_ after 2.5sec.
So if an app loads a bunch of stuff between the "window.load" and the "_moz-app-visually-complete" events, it'll be competing with WindowManager work.