Last Comment Bug 947390 - [meta][project] Improve page load performance
: [meta][project] Improve page load performance
Status: NEW
: meta, perf
Product: Firefox for Android
Classification: Client Software
Component: General (show other bugs)
: unspecified
: All Android
-- normal with 8 votes (vote)
: ---
Assigned To: Nobody; OK to take it and work on it
:
: Sebastian Kaspari (:sebastian)
Mentors:
Depends on: 899424 916968 924712 941155 945408 947067 971232 971239 899425 917896 943475 945779 947393 948565 949104 949639 949913 951959 952473 952799 952878 952894 955870 957037 958087 963505 964307 966573 969835 970604
Blocks:
  Show dependency treegraph
 
Reported: 2013-12-06 12:09 PST by Michael Comella (:mcomella) [not active on fennec/Bugzilla: contact me via IRC]
Modified: 2016-01-10 04:46 PST (History)
15 users (show)
See Also:
Crash Signature:
(edit)
QA Whiteboard:
Iteration: ---
Points: ---
Has Regression Range: ---
Has STR: ---


Attachments

Description User image Michael Comella (:mcomella) [not active on fennec/Bugzilla: contact me via IRC] 2013-12-06 12:09:00 PST
In the 12/4 weekly meeting [1], we discussed a gradual performance regression in page load times over the past few months, discovered by comparisons to Chrome and Eideticker.

Currently, we can track page load time with the full product by loading a page and comparing the result with older versions. This is a good start but we should also create a method of determining how much page load performance is impacted by Android-specific code so we can find regressions there.

For example, one suggestion is to build Gecko into a standalone app (perhaps using GeckoView) that will startup and immediately load some web page, and compare this to our full app results (ex: Eideticker).

There may also be work to be done on profiling and seeing where we might be slow, perhaps independently from whether we have regressed or not. For example, there may be some inefficiencies where code is called multiple times ([2], [3]).

Finally, there is work to be done on page load performance that is not a direct regression, such as predictive lookup and prefetching (see [4] and its blocking bugs).

[1]: https://wiki.mozilla.org/Mobile/Notes/04-Dec-2013
[2]: https://bugzilla.mozilla.org/show_bug.cgi?id=924712
[3]: https://bugzilla.mozilla.org/show_bug.cgi?id=941844
[4]: https://bugzilla.mozilla.org/show_bug.cgi?id=943475
Comment 1 User image Mark Finkle (:mfinkle) (use needinfo?) 2013-12-17 16:48:03 PST
Adding a few of the redirect bugs that make GlobalHistory do less work during a pageload. Less writing to the DB means more time for networking.

Also adding the Seer pageload regression.
Comment 2 User image Richard Newman [:rnewman] 2013-12-22 19:37:27 PST
I think this bug is/has morphed into the tracking bug for the page load performance project, so gently recasting it as such, under the assumption that some of "improve" will be "find and fix regressions".

mfinkle, do we still not track meta bugs? Is the weekly meeting enough spinning for these project 'plates'?

Note You need to log in before you can comment on or make changes to this bug.