| Age | Commit message (Collapse) | Author |
|---|
|
|
|
- Move the slide editing test into the main runWorkload instead of
setup.
- On some devices the folder picker has changed layout so add support for
navigating.
- Add support for differently capitalized splash buttons.
- Add workaround for adding a new slide if click the button doesn't work
the first time.
|
|
- Fix formatting
- Skip teardown automation if elements are not present on some devices
instead of failing the workload.
- Give extra time for start button to appear as some devices can be slow
to load.
|
|
Add additional check for text based directions button as id can be
missing on some devices and allow for skipping the view steps stage for
large screen devices which do not require this step.
|
|
On some devices the entire results page fits on one screen and does not
present a scrollable element, therefore only attempt to scroll if
available.
|
|
|
|
Previously the configuration of the tests was performed in the run stage
instead of the setup.
|
|
The X coordinate was miscalculated when attempting to load the test
selection menu.
|
|
|
|
On some devices the score string obtained can contain extra characters.
Only use the numerical values from the score when converting, otherwise
if not found set the result to 'NaN'.
|
|
v4.3.1 has made a minor change to the run cpu benchmark element.
Refactoring to support both the new and previous elements.
|
|
By clearing the application data each time the workload is run this
forces the required assets to be re-installed each time. As the
workload is not affected by persistent state do not perform the
clearing.
|
|
|
|
Add check for a differently worded popup informing that assets are
missing.
|
|
|
|
Update workloads that use uiautomator and can display a warning about
using an old version of the app to dismiss the popup if present.
|
|
While attempting to run vellamo on Android Q, a popup warning with
the message, "This app was built for an older version of Android and may not
work properly. Try checking for updates, or contact the developer." would
appear, causing the workload to halt.
Close the popup warning before dismissing EULA and executing the remaining
steps to run vellamo.
Tested with vellamo apk version 3.2.4.
Signed-off-by: Lisa Nguyen <lisa.nguyen@linaro.org>
|
|
Previously the LogcatRunMonitor left the logcat process running in the
background causing issues with concurrent accesses. Now ensure the thread
terminates correctly.
|
|
Ensure output is encoded correctly if running with python 3
|
|
For workloads that support Linux targets do not use
`uninstall_executable` as this is not available, instead use `uninstall` as
other targets should be able to determine the appropriate uninstallation
method.
|
|
Previously the workload assumed that all the scores were visible on a
single screen however for devices with smaller displays the results need
to scrolled.
|
|
A new popup appears when running pcmark on android Q that complains
about the app being built for an older version of android.
Since this popup will be temporary, the fix has to make sure not to
break in the future when this popup disappears or when the test is ran
on a compatible version of android.
To achieve this, we attempt to dismiss the popup and if we timeout we
silently carry on with the test assuming no popup will appear.
Signed-off-by: Qais Yousef <qais.yousef@arm.com>
|
|
Creating a new workload to execute the following tests on GFXBench.
* Car Chase
* Car Chase Offscreen
* Manhattan 3.1
* 1080p Manhattan 3.1 Offscreen
* 1440p Manhattan 3.1 Offscreen
* Tessellation
* Tessellation Offscreen
|
|
Updating the googlephotos workload to work with app version 4.0.0.212659618
|
|
Some devices throw errors if too many browser tabs are open. Have
added a method to close down the tabs in the teardown function.
|
|
Leftover code from WA2 meant that logcat was cleared on the device by
the workload directly instead of using devlib, this caused issues if logcat was
still being cleared from other areas of the code.
|
|
When running jankbench invalid bytes can be read from the device causing
decoding in the monitor to fail, now replace any invalid sequences.
|
|
|
|
|
|
Ensure that the results file is decoded when using python3.
|
|
|
|
|
|
On some devices running Android 8.1 the start benchmark button was
failing to be clicked, this is a workaround to click on the coordinates
of the button instead of the UiObject iteslf.
|
|
Implement fixes for the most recent pylint version.
|
|
|
|
|
|
|
|
- Add copyright headers to files that did not already have them
- Update the year to the last year the file was modified in files with
existing headers.
|
|
|
|
Instead of using the os to determine whether the screen state should
be checked, use target flag to see if running inside a container.
|
|
Mongoperf currently has a non-ASCII character in the file that causes a failure.
This commit rewords the description to remove the incriminating character.
|
|
Fix incorrect method name.
|
|
We cannot currently determine the screen state from cannot android
containter so only check on android systems.
|
|
|
|
Add a workload that runs mongoperf benchmark that measures I/O
performance on a MongoDB server. This workload assumes that mongoperf
is already installed.
|
|
Lengthen timeout for gemm and conv tests to be the same as for sparse
test. While the former two usually take a lot less time, their execution
time will vary significantly depending on the target and the runtime
environment (e.g. cpu frequencies might be forced to lowest values).
|
|
urllib2 does not exist in Python 3, and its methods have been moved into
urllib.request. Use future library to create aliases that work across
both 2 and 3.
|
|
Add missing blank lines around the URL indent block.
|
|
Add a workload for lat_mem_rd and bw_mem tests of LMbench benchmark, a suite of
portable ANSI/C microbenchmarks for UNIX/POSIX.
This is a port of corresponding workload from WA2.
|
|
|
