Modify the JetNews workload to use the new jank testing classes. This allows
simplification of the code and also makes it more straightforward to implement
new workloads that also need to gather frame metrics.
Create two new classes to deal with workloads for which we want to
collect frame metrics/jank percentages.
ApkUiautoJankTestWorkload inherits from ApkUiautoWorkload and instantiates
a UiAutomatorJankTestGUI gui member.
UiAutomatorJankTestGUI (inherits from UiAutomatorGUI) handles most of the
new logic for jank tests, including using a regular expression to parse the
final output.
This patch adds the JetNews jank-testing workload. This is accomplished
through the uiauto helper library and the jank test classes.
This workload requires a JetNews app APK to be available. We plan to
make it available through the workload-automation-assets repo.
At the end of the run, users should end up with a json file containing all
the frame/jank metrics.
There are 3 parameters for this workload:
- tests: Specifies which of the 3 available tests to run (default is to run
all of them)
- flingspeed: The speed of the fling interactions.
- repeat: How many times each of the selected tests is to be executed in a
single measuring session.
Jank testing involves using a different library (JankTestBase) that our
testing class needs to extend. Since we already make the base testing
class UiAutomation extend BaseUiAutomation, we cannot use the UiAutomation
class for our tests.
The uiauto_class property is hardcoded to "UiAutomation" though, so this
patch adds a uiauto_jank_class property for the Jank test class.
While at it, we also make the uiauto_runner configurable (currently hardcoded
to the old deprecated android.support version of it) so that our workloads
can pass whatever runner they need to use.
In terms of stages for testing, their current names are the following:
stages = ['setup', 'runWorkload', 'extractResults', 'teardown']
Those won't work for jank tests. For some reason the jank test library wants
to see the testing function name contain or start with a "test" string
pattern on it.
This behavior doesn't seem to be documented, but at least with the current
version of the JankTestBase library we have this restriction.
To make things work despite this restriction, I've added separate jank
test stages that can be provided by the workload. The generic workload
code will use these stages to generate the instrumentation invocation
command, and the provided jank test stages (essentially jank test
functions) will be invoked one by one.
Finally, we make the UiAutomatorGUI's _execute method return the output
of the execution so the workload can freely process and pattern-match
the results, like the UiBenchJankTests workload.
Currently the uiauto library is built on top of the deprecated android.support
libraries. Create an updated uiauto library (uiauto-androidx) using androidx
and keep it alongside the old one for compatibility reasons.
As we update workload-automation going forward, we should update the workloads
to use the androidx version of uiauto.
This has been tested with the latest gradle version available at the time
(8.8).
When speedometer is running on a Chrome package on Android 14, a pop-up
window was showing on the screen. Chrome preferences file is modified
to dismiss the window.
The key for 'include#' can now be either a scalar or a list.
A scalar triggers the same behaviour as before.
If the value is a list it must be a list of scalars (filepaths).
The paths will be loaded and merged in order, and finally the resulting
dict is included into the current scope.
- ``load_struct_from_yaml()`` has been moved to devlib [1].
- ``LoadSyntaxError()`` is already implemented in devlib.
- Remove ``load_struct_from_file()`` and ``RAND_MOD_NAME_LEN`` since
they are not used at all.
[1] https://github.com/ARM-software/devlib/commit/591825834028
Signed-off-by: Metin Kaya <metin.kaya@arm.com>
Use importlib.import_module instead of __import__ as per Python doc
recommendation.
This will also fix the case where the class is in a
package's submodule (since __import__ returns the top-level package),
e.g. "foo.bar.Class".
Introduce 'unlock_screen' option in order to help in automating Android
tests by unlocking device screen automatically. Surely this works only
if no passcode is set.
'unlock_screen' option implicitly requires turning on the screen. IOW,
it will override value of 'screen_on' option.
'diagonal', 'vertical' and 'horizontal' are valid values for
'unlock_screen' option as of now.
Note that this patch depends on
https://github.com/ARM-software/devlib/pull/659 in devlib repo.
Signed-off-by: Metin Kaya <metin.kaya@arm.com>
"--user <USER_ID" (current user: 0) option is added to
activity manager (am) command because of "Invalid userId" command.
Tested with other benchmarks (geekbench) as well.
Add an instrument that uses devlib's PerfettoCollector to collect a
Perfetto trace during the execution of a workload.
The instrument takes a path to a Perfetto config file which specifies
how Perfetto should be configured for the tracing.
Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>
Bromite is a fork of Chromium that's easily available for Android. Apart
from small changes it works the same as Chromium and works with this
speedometer workload. Add it to the 'package_names' list to allow using
it as an option.
https://www.bromite.org/
Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>
Update rt-app binaries to the latest version of the "lisa" branch in
douglas-raillard-arm GitHub fork. This tracks the upstream master branch
with a number of critical patches required notably to work with uclamp.
Allow optionally setting the top level ftrace buffer size separately
from the devlib buffer size. The parameter will be passed to the devlib
FtraceCollector and take effect there.
Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>
WA_PLUGIN_PATHS currently contains a list of filesystem paths to scan
for plugins. This is appropriate for end-user plugins, but this is
problematic for plugins distributed by a 3rd party, such as a plugin
installed from PyPI.
In those cases, the path to the sources is unknown and typically depends
on the specify Python version, local setup etc. What is constant is
Python name of the package, e.g. "lisa.wa.plugins".
Extend the input allowed in WA_PLUGIN_PATHS by trying to load entries as
a Python package name if:
* There is no filesystem path with that name
* The entry is a "relative path" (from an fs point of view)
Set the view parameter so that the FPS instrument can collect frame data
from the workload.
Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>
Rework the uibenchjanktests workload to allow specifying a list of
subtests. The activity will be re-launched for each provided subtest. If
none are specified, all available tests will be run in alphabetical order.
The workload output will now include metrics with their respective test
names as classifiers.
Add a 'full' parameter to revert back to the old default 'full run'
behaviour with restarts between subtests.
Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>
Add a workload for the Dr Arm demo app. Includes functionality for
automatically pulling the ADPF FPS report file from the target if one
was generated by the app.
Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>
The simpleperf included with Android 13 now does not show the percentage
when no counter multiplexing took place. This causes the perf instrument
to crash when processing the output. This fix checks whether the percentage
exists before trying to extract it.
Signed-off-by: Kajetan Puchalski <kajetan.puchalski@arm.com>
In some workloads it is beneficial to be able to provide arguments
when launching the required APK. Add a new `arpk_arguments` property
to allow a workload to provide a dict of parameter names and values
that should be used when launching the APK.
The python types of the parameters are used to determine the data
type provided to the APK. Currently supported types are string, bool,
int and float.
Add support for Geekbench 6 as a workload on Android.
This commit adds 6.*.* as a valid version for the Geekbench workload and
updates the UIAuto apk accordingly.
It also refactors the update_result function seeing as the one
originally used for GB4 can now be used for 4, 5 and 6 and so it makes
more sense to treat it as a 'generic' update_result function. The
functionality should stay the same.
Backwards compatibility with GB2 & GB3 should be maintained.
Implements support for dynamically loading additional RuntimeConfig
and associated RuntimeParameter that are defined in a plugin.
Currently, the various RuntimeConfig's are hard coded in a list within
WA. This patch extends RuntimeParameterManager to use PluginLoader to
load RuntimeConfig classes and append to the hard coded list.
The implementation, as written, does not allow loading RuntimeConfig
from a plugin if it has the same name as one of the hard coded
RuntimeConfig. This is meant to prevent conflicts and unexpected
behavior.
Currently, when processing the output of 'simpleperf stat', wa does not
skip the header and tries to process part of it as a number, leading
to type errors. This change skips the header (line starting with '#').
Furthermore, some events (e.g. cpu-clock or task-clock) include "(ms)"
in their count value and are floats instead of integers. Because of
this, when either of those is included, processing metrics fails due to
assuming every metric is an integer. Then another error happens when the
code tries to split the line on '(' assuming that there's only one set
of those around the percentage.
This change removes "(ms)" from the line
before it's processed and properly determines whether 'count' is an
integer or a float before attempting to convert it.
The non corporate version of geekbench5 didn't work although the code
had everything needed, except for a number of tiny required tweaks:
1. Add '5' in the supported versions in __init__.py
2. Fix the name of the android package in__init__.py and
UiAutomation.java
3. Improve handling of minorVersion to fix potential exception when we
don't specify the minorVersion number in the yaml file. Launching
geekbench5 works fine when it's the only one installed. But if you
have multiple versions, then using the version string in the yaml
agenda didn't like specifying '5' as the version and threw exception
out of bound because we assume '5.X' as input. No reason I'm aware of
to force support for a specific version of geekbench5. So keep it
relaxed until we know for sure it breaks with a specific version.
Signed-off-by: Qais Yousef <qais.yousef@arm.com>
Which works as the corporate version except for a different in package
name and a set of Fixed Time Tests that don't exist on free version.
Only support v 4.X and v5.X as that's what's available.
Note there's a clash with glbenchmark package name. glbenchmark is an
ancient version provided by the same developers but was superseded by
gfxbench. The version checks in both workloads should ensure we get the
right one in the unlikely case both are installed.
Signed-off-by: Qais Yousef <qais.yousef@arm.com>
