| Age | Commit message (Collapse) | Author |
|---|
|
Test: manual
Change-Id: Id81a1a2ef704da20b99533a432199bdc8150ed51
|
|
cpufreq_go used the plain adb, this changes it to use $adb,
which is set to adb -s if -s was passed, allowing the go test
to run when multiple devices are connected.
Bug: 72652287
Test: manual
Change-Id: Iffd658e9f1f99cbe992a5e1a5a07adc70d334653
|
|
am: c500bdd557
Change-Id: Icb8574c2f32ec41a126e550bcac7c63a03487ae2
|
|
am: aa317f0c4c
Change-Id: I4792be67f2a33c61f0e740ba9093c06df4613921
|
|
|
|
* Remove unused variables.
* Fix redefined macro warnings.
* Fix sign-compare warnings.
* Fix 'return false' in main().
* Keep existing warnings to fix later.
Bug: 66996870
Test: build with WITH_TIDY=1
Change-Id: Ib92ef5297693595fb84ed4f8e61665bda4cee312
|
|
Device utilization computation converted the values to hz, assuming the
device is 100 HZ, which is not necessarily the case. This change converts
everything to ms. Device utilization computation now works on marlin (P).
Test: Test app-launcher and make sure device util% reported is correct.
Change-Id: Id1dfe7e21fbd07d89fbe8a87ec49f76b17f7429e
Signed-off-by: Mohan Srinivasan <srmohan@google.com>
|
|
Add support for Go devices.
Test: Start app-launcher on Go and verify that the data reported looks
sane.
Change-Id: I1bc0bd23e7d807444f0ac35d7c18c8206bb7c12e
Signed-off-by: Mohan Srinivasan <srmohan@google.com>
|
|
There's some additional setup needed to make sure the devices are all in
the same state, so I added instructions to the README.
I also changed the system_stats functions to use variables to store
string outputs instead of files to avoid collision when running multiple
instances of the script from the same directory.
Finally, I added cleanup at the end to remove the files created in the
process, and some general code cleanup throughout the script.
Bug: 62536649
Test: manual test on several devices
Change-Id: I843a1cefcf6bd11f28e513d735b3bf313442efd2
|
|
The script previously did not configure any framework settings for this
script, so I added commands to disable any features that could affect
the performance numbers like wifi, NFC, screen rotation. It also clears
notifications and starts the device on the home screen.
It's not possible prior to this commit to run the app-launcher on a
host with multiple devices attached to adb, so I added a '-s <serial
number' option that allows the user to specify a specific device based
on its serial number.
I also changed the spacing of the script to be all spaces instead of
mixed tabs and spaces.
Test: Manually ran script on a marlin device
Change-Id: I82c621fe280033c55ba38bfdf95836fd95f5ed58
|
|
Use TotalTime instead of ThisTime (both reported by am) as our
activity launch latency metric. TotalTime is more representative
of user perceived latency. Also add back YouTube to the list of
apps we launch test by default.
Test: Time app launches manually for the various apps, and check
that TotalTime better represents the launch latency.
Change-Id: I127a22ae409ee0141d847e711e09f7873dbb420f
Signed-off-by: Mohan Srinivasan <srmohan@google.com>
|
|
app-launcher is an automated tool that automates the launching of
selected Apps (hundreds of times each) and collects system metrics
like CPU util, storage device util, launch times, context switches,
branch mispredicts etc. This is very useful for A/B compares of system
changes in how they impact App Launch. This code was under
google_experimental and was used by various members who found it very
useful, so moving it under system/extras (under userdebug).
Test: I (and others) have used app launcher on multiple occasions over
several months. And comparison of metrics reported by app-launcher are
consistent with metrics reported by other tools, for these tests.
Change-Id: I8b469b4165035a958b155cfffc93341cddcc444c
Signed-off-by: Mohan Srinivasan <srmohan@google.com>
|
