Start a capture session to profile data from your application in real time. When the capture session finishes, Streamline automatically opens a report for you to analyze later.
For instructions on how to capture data locally on a target, and for more target-specific information, see the Streamline target setup guides.
Set up your Linux or Android target as described in the relevant Streamline target setup guide.
In the Start view, select your target device type. Then select your device from the list of detected targets, or enter the address of your target.
Note
By default, gatord uses port 8080 to communicate with the host machine. To use a different port, append a colon followed by the port number to the target address. For example, to use port 5050 with target address 10.99.28.54, enter 10.99.28.54:5050 in the target address field.
Android users only, select the package you want to profile from the list of packages available on the selected device.
Note
To profile non-debuggable packages in the list, you must be running a development device with root access.
Applications must have at least one MAIN or launchable activity.
Optionally, enter arguments for activities that require additional configuration, or for activities that cannot be listed in the table:
Select the Override activity check box, and enter a suitable name for the activity.
In the Arguments text box, enter the arguments that you want to run on the selected package.
You can enter multiple arguments passed on a single line. See valid argument options in https://developer.android.com/tools/adb#IntentSpec.
If the arguments are valid, when you click Start capture, the arguments are stored in history so that you can select them again from a drop-down menu. A maximum of 10 arguments are stored before they are overwritten.
TCP users only, optionally enter the details for any command you want to run on the application. See Run a command on the target.
Select the counter template that you want to use to review performance of your CPU and GPU:
To use the most appropriate counter template for your Arm GPU, select the Capture Arm GPU profile checkbox.
To use the default counter template, which captures basic information such as CPU and memory usage, but no GPU data, clear the Capture Arm GPU profile checkbox.
Note
If your device does not contain an Arm GPU, the Capture Arm GPU profile checkbox is disabled.
To use a different counter template, select the Use advanced mode checkbox.
Click Select Counters.
Click Add counters from a template
to see a list of available templates.
Select a counter template appropriate for the GPU in your device, then Save your changes.
The number of counters in the template that your device supports is shown next to each template. For example, here, 101 of the 101 available counters in the Arm® Mali™ template are supported in the connected device. Streamline notifies you if the target device does not support all the counters that are defined in the selected template.
Alternatively, to import an existing counter template, click Install and Load.
Optionally, click Capture Settings to set more capture options, including the sample rate and the capture duration (by default unlimited).
Click Start capture.
Browse to a location on the host and enter the name of the capture file.
Run the application scenario that you want to profile on your target device.
In the Capture Control view:
To stop the capture session and create a performance report, click Stop capture and analyze
.
To stop the capture session and discard the collected data, click Stop capture and discard
.
If Stop capture and analyze
is grayed out, Discard data has been selected in the Capture Settings dialog box. To generate a performance report, clear Discard data in the Capture Settings dialog box, click Save to close the dialog box, then click Start capture.
When data analysis completes, the application terminates on the device and the report automatically opens in the Timeline view.
See tutorial Android performance triage with Streamline on the Arm Developer website.