915123 - [1.2] Regression in SMS Scroll FPS (compared to [1.1])

Status: RESOLVED WORKSFORME

(Firefox OS Graveyard :: Gaia::SMS, defect, P1)

Description

[Mandyam Vikram]

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/29.0.1547.66 Safari/537.36

Steps to reproduce:

Based on high speed camera measurements (on a QRD 7x27 device), there seems to be a degradation in UX scroll performance of the SMS App.



Actual results:

Measurements on 1.2:
42 FPS 
Measurements on 1.1:
52 FPS

Comment 2

[Hubert Figuiere [:hub]]

Is it scrolling for the threads or of the thread content? What kind of workload do you have?

Comment 3

[Mandyam Vikram]

Please refer to https://bugzilla.mozilla.org/show_bug.cgi?id=915088#c4

Comment 4

[Hubert Figuiere [:hub]]

b2gperf has an experimental scrollfps test for contacts.

https://github.com/davehunt/b2gperf/tree/0.9

It is not yet merged and run because of some issues with Hamachi but seems to work on inari. I'll put up numbers rsn.

Comment 5

[Hubert Figuiere [:hub]]

I did use a |make reference-workload-x-heavy| that put 2000 messages in the SMS app.

I run the 0.9 tree for b2gperf against 1.2 from yesterday and 1.1 downloaded from pvtbuilds:
https://pvtbuilds.mozilla.org/pvt/mozilla.org/b2gotoro/nightly/mozilla-b2g18-inari-eng/2013/09/2013-09-27-04-12-01/

Here are the results:

1.2

2013-09-27 13:40:24 | INFO | Results for messages, fps: median:57, mean:57, std: 0, max:58.408768889, min:57.5074460821, all:57.7714046115,57.8596498308,58.408768889,57.654966467,57.902688565,57.9715612768,58.0704587396,57.8932438423,57.6500350012,57.7534355924,57.7389548203,57.5074460821,57.9476488844,58.0693073121,57.9244377379,58.0512536763,58.0978961797,58.0304565962,58.1670409516,58.0455769978,58.1276436306,57.9984741973,58.2185736527,58.1578896697,57.7539107223,57.9228313641,57.6046137182,57.9284469649,57.9719265294,57.6336620827

1.1

2013-09-27 16:04:17 | INFO | Results for messages, fps: median:54, mean:52, std: 7, max:61.9364989576, min:33.6162913282, all:61.0839100661,58.0778779364,55.7171727296,38.1236761084,40.8738342706,56.3594704991,61.5637778179,55.530236431,58.2520800779,61.9364989576,52.2397196071,52.8063970557,48.3993251932,41.9451810419,46.4890227314,60.2567384272,54.6268099354,49.7213824903,33.6162913282,36.7457804303,58.9123612073,59.420639149,43.540065121,58.9611259375,54.3270917103,47.8111953164,50.2319877368,53.7045546816,51.8966582157,58.2273866159

In short, 1.2 has a better FPS than 1.1

This test has been run on Inari.

Comment 6

[Hubert Figuiere [:hub]]

I wonder if this wasn't fixed by itself as part of the other work?

I don't have your high-speed camera setup, but our b2gperf show that it is as good is not better.

Do you mind retesting?

Comment 7

[Mandyam Vikram]

I've sent recent results to Sandip.

Comment 8

[Hubert Figuiere [:hub]]

Latest measurements show that we are better than last time.
Issue got solved with other work on m-c.

Thanks.