Now that you have seen an example Streamline capture, you can use it with your own application.
The application must be compiled with debug enabled, as well as additional options to facilitate call stack unwinding by Streamline.
The Target setup guide for Android provides appropriate instructions for C++ or Java applications.
These settings are enabled as default within
Unity
when
Development Build
is selected in Build settings
.
You can optionally provide annotations to your code which will be displayed in the Streamline report, to assist readability of the report.
Arm provides open-source projects that can be used by application developers as part of their development.
Ensure that
Developer Mode
is enabled, then enable USB Debugging
under Settings > Developer options
.
Connect the device to the host through USB and approve the debug connection when prompted.
If the connection is successful, running the adb devices
command on the host returns the ID of your device, and you can run adb shell
.
Install the debuggable application.
To set Unity
applications to be debuggable, enable Development Build
in Build Settings
.
Unity Build Settings
In the Streamline Start
view, select Android (adb)
as your device type, then select your device from the list of detected devices.
Connect to the device
This will install the gatord
daemon and connect to the device.
Wait for the list of available packages to populate, then select the one you wish to profile from the list of available packages on the selected device.
Select an appropriate counter template for the GPU present in your device.
Navigate to Window
> Preferences
> Data Locations
, and configure location to store data. New reports will be created in the topmost folder specified.
Click on Start capture
to start capturing profile data from the target.
Start the application on the device, and interact as desired for the profiling run you wish to do.
When satisfied, simply click on Stop capture
in upper-right of screen.
Stop Capture
Streamline will stop capturing data, remove the daemon, and process the captured data.
Double-click on the capture to display in the timeline.