Profile executable and linkable format binaries on Android

You can profile Executable and Linkable Format (ELF) binaries using the command-line, or the Streamline GUI.

Before you begin

  • Ensure your application is compiled with symbols and debug information.

  • Ensure your Android device is connected to your host computer.

  • Ensure gatord and the application binary are located in /data/local/tmp on the Android device. To install gatord, see Build gatord yourself.

About this task

This task explains how to profile ELF binaries using the command-line. If you want to profile ELF binaries using the Streamline GUI, see Run a command on the target in the Arm Streamline User Guide.

Procedure

  1. To profile your ELF binaries, use adb to run gatord in local capture mode, and instruct it to run the application that you want to profile on the Android device:

    adb shell

cd /data/local/tmp

./gatord -o capture.apc -A <elf_application>

    Note

    To change the counter configuration, use either the -C argument or the --config-xml argument. See gatord command-line options.

    gatord generates a profile on your Android device at /data/local/tmp/capture.apc.

  2. Exit the shell on the Android device, then move the profile to the Streamline data directory:

    exit

adb pull /data/local/tmp/capture.apc <streamline-data-directory>

    Note

    The Streamline data directory is configured in Window → Preferences → Streamline → Data Locations.

  3. Open Streamline on your host computer.

    Your profile is available in the Streamline Data tab.

  4. Add the ELF binary file before you analyze a profile of your application in Streamline.

    1. Open Streamline.

    2. In the Streamline Data tab, right-click the profile that you want to analyze.

    3. Click Analyze ....

    4. In the Program Images tab, click Add image.

    5. Select the ELF files that contain the debug symbols for the application and libraries.

    6. Click Analyze.

    In the Call Paths and Functions views, the Processes and Functions are named appropriately.

Next steps

You can now analyze your profile in Streamline. See Analyze your capture in the Arm Streamline User Guide.