Closed
Bug 1094011
Opened 10 years ago
Closed 7 years ago
Profile template app to find the baseline overhead on multi-core device
Categories
(Firefox OS Graveyard :: Performance, defect)
Firefox OS Graveyard
Performance
Tracking
(tracking-b2g:backlog)
RESOLVED
WONTFIX
tracking-b2g | backlog |
People
(Reporter: kanru, Unassigned)
References
Details
Profile the template app to find the startup time overhead and what could be improved on multi-core device.
We need to compare the behavior on 1, 2, or 4 cpu cores.
Updated•10 years ago
|
Blocks: AppStartup
Updated•10 years ago
|
No longer blocks: B2G-Multicore
Comment 1•10 years ago
|
||
Instead of the template app, this [1] seems a more realistic one. See the original post [2].
[1] https://github.com/mozilla-b2g/gaia/tree/master/dev_apps/test-startup-limit
[2] https://groups.google.com/forum/#!msg/mozilla.dev.gaia/1rEMVKeUfIw/mFsG5mgJbnoJ
Comment 2•10 years ago
|
||
This is the test result on my Flame 319MB (v118D) today, with versions:
Gaia: ad1ef4b324faf0e8
Gecko: https://hg.mozilla.org/mozilla-central/rev/c8c4a2746aac
ting@sweet:~/w/fx/os/flame/gaia$ RUNS=20 APP=test-startup-limit node tests/raptor/launch_test
[Raptor] Preparing to start testing...
[Cold Launch: Test Startup Limit] Priming application
[Cold Launch: Test Startup Limit] Starting run 1
...
[Cold Launch: Test Startup Limit] Run 20 complete
Metric Mean Median Min Max StdDev p95
-------------------------------- ------- ------- ------- ------- ------ -------
coldlaunch.navigationLoaded 230.100 229.000 193.000 313.000 24.738 284.500
coldlaunch.navigationInteractive 266.100 264.500 225.000 361.000 28.053 324.500
coldlaunch.visuallyLoaded 806.750 781.500 735.000 975.000 63.796 934.000
coldlaunch.contentInteractive 807.100 782.000 735.000 976.000 63.822 934.500
coldlaunch.fullyLoaded 808.250 783.000 736.000 980.000 64.389 937.000
coldlaunch.pss 8.245 8.200 8.200 8.300 0.050 8.300
coldlaunch.uss 0.000 0.000 0.000 0.000 0.000 n/a
coldlaunch.rss 13.365 13.400 13.300 13.400 0.048 13.400
[Cold Launch: Test Startup Limit] Testing complete
But on Raptor [1], the same version has visuallyLoaded over 1s. Eli, do you know what could make the difference?
[1] http://raptor.mozilla.org/#/dashboard/script/mark.js?device=flame-kk&branch=master&memory=319&suite=coldlaunch&to=1431646696471&from=1431641136819&panelId=13&fullscreen
Flags: needinfo?(eperelman)
Comment 3•10 years ago
|
||
I know the Jenkins devices use reference workloads and may have other environmental differences. When was the last time you did `make raptor`?
Flags: needinfo?(eperelman)
Comment 4•10 years ago
|
||
Right before testing.
Comment 5•10 years ago
|
||
(In reply to :Eli Perelman from comment #3)
> I know the Jenkins devices use reference workloads and may have other
> environmental differences.
The test right after |./flash.sh|, |make raptor| and |make referece-workload-medium|, I got this:
ting@sweet:~/w/fx/os/flame/gaia$ RUNS=20 APP=test-startup-limit node tests/raptor/launch_test
[Raptor] Preparing to start testing...
[Cold Launch: Test Startup Limit] Priming application
...
Metric Mean Median Min Max StdDev p95
-------------------------------- -------- ------- ------- -------- ------- --------
coldlaunch.navigationLoaded 676.550 241.500 191.000 2533.000 882.461 2520.000
coldlaunch.navigationInteractive 715.650 285.500 226.000 2571.000 878.097 2554.000
coldlaunch.visuallyLoaded 1196.600 789.000 722.000 2937.000 810.044 2908.500
coldlaunch.contentInteractive 1197.000 789.500 722.000 2937.000 809.962 2908.500
coldlaunch.fullyLoaded 1197.850 790.000 723.000 2938.000 810.051 2909.500
coldlaunch.rss 14.490 13.300 13.200 19.300 2.405 19.300
coldlaunch.uss 0.000 0.000 0.000 0.000 0.000 n/a
coldlaunch.pss 9.770 8.750 8.700 13.900 2.065 13.900
[Cold Launch: Test Startup Limit] Testing complete
From b2g-info, Preallocated process was shown at 5th run, probably due to busy indexedDB.
Updated•9 years ago
|
No longer blocks: AppStartup
Comment 7•7 years ago
|
||
Firefox OS is not being worked on
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → WONTFIX
You need to log in
before you can comment on or make changes to this bug.
Description
•