Investigate screen noise in android power measurements
Categories
(Testing :: Raptor, task, P3)
Tracking
(Not tracked)
People
(Reporter: sparky, Assigned: sparky)
References
Details
Attachments
(3 files)
screen_power_vs_temperature.png
This bug is for investigating and reducing noise in android screen power measurements. Currently, we have two possibilities of what is introducing this noise, (i) device temperature, and (ii) the device being tested.
See bug 1558189 for some work on noise caused by temperature, and bug 1558191 for device-dependent noise.
|
Assignee | |
Comment 1•5 years ago
•
|
||
Device battery temperature has no correlation with screen power so it cannot be the culprit.
This graph is for 175 historic data points on android-p2-8-android-aarch64/raptor-speedometer-power from May 19, 2019 (after screen power was increased to 50% of the maximum). Pixel 2 is the only device which exhibits this noise.
The code and data (downloaded by the script) for this analysis can be found here: https://github.com/gmierz/android-screen-noise-investigation
|
|
Assignee | |
Comment 2•5 years ago
|
||
Device name is correlated to screen power. So this noise is occurring because each of the devices has two different power ranges for the screen power.
It seems like it could be coming from the adaptive brightness setting if it didn't get turned off, or from the screen hardware itself.
I have a patch for disabling adaptive brightness that I will be submitting for review shortly.
|
|
Assignee | |
Comment 3•5 years ago
|
||
Closing this now since the try run here shows that the bimodality is gone when using the patch from bug 1572862: https://treeherder.mozilla.org/#/jobs?repo=try&revision=d6a94865d8d3009127fba115cff7963b12929f1d
Thanks, Greg; verified FIXED. For posterity, see attached screenshot of new graph plots from https://treeherder.mozilla.org/perf.html#/graphs?series=try,1832494,1,10&series=try,1832499,1,10&selected=try,1832499,530228,887261142,10.
Description
•