Build and run an example application to obtain Arm GPU configuration information

You may want to adjust application settings at runtime to deliver the best performance from device hardware. Querying the hardware configuration is the first step in making decisions which take advantage of available hardware features.

libGPUInfo is a C++ library which can be integrated into applications to easily gather information about Arm GPU hardware.

You can build an example application and run it on an Android device with an Arm GPU to read the hardware configuration information. This information identifies available features and performance levels.

Download and install the Android NDK

The libGPUInfo library and example application is written in C++. You can build it with the Android NDK (Native Development Kit).

You can install the Android NDK on an x86_64 Ubuntu or Debian machine using:

    

        
        
            sudo apt update
sudo apt install cmake google-android-ndk-installer -y
        
    

There are other ways to install the Android NDK and additional operating systems you can use. Refer to NDK Downloads and Install NDK and CMake for more options.

Build the library and application

Retrieve the software using the git command below and change to the repository directory:

    

        
        
            git clone https://github.com/ARM-software/libGPUInfo.git
cd libGPUInfo
        
    

Set the Android NDK environment variable to the location where the NDK is installed:

    

        
        
            export ANDROID_NDK_HOME=/usr/lib/android-ndk
        
    

Build the application by running the script:

    

        
        
            bash ./android_build.sh
        
    

The android_build.sh script recognizes Release and Debug as arguments. The default is Release if no argument is supplied. Add Debug as a command line argument if you want to build a debug version.

Confirm the file exists using:

    

        
        
            ls bin/arm_gpuinfo
        
    

If there are no build errors and the file name is printed, the application is ready to use.

Run the example application on an Android device

To run the application on an Android device you need a computer with adb (Android Debug Bridge) installed.

You can connect to the Android device using USB or Wi-Fi. You can use adb to copy arm_gpuinfo from your computer to the Android device.

If the computer you will use for adb is not the same one you used to build the application, you can copy the arm_gpuinfo application from the build computer to the adb computer.

1. Install adb from Android Platform Tools

To install adb on Ubuntu or Debian run:

    

        
        
            sudo apt-get install android-sdk-platform-tools -y
        
    

Refer to the SDK Platform Tools release notes for installation information about other operating systems.

2. Enable developer options

To enable debugging and other developer options visit the device settings on your Android device, find the Build Number under About phone or About tablet, and tap the Build number 7 times until you see a message You are a developer!

A new menu for Developer options is now visible and you can enable USB debugging and Wireless debugging.

The instructions below use USB debugging. For more information about wireless debugging refer to the ADB developer documentation . You will use the adb pair command and a PIN to connect to the Android device over Wi-Fi.

3. Connect to Android

Connect a USB cable between your computer with adb installed and the Android device. A prompt saying Allow USB Debugging? may appear on your device. If so, press OK.

Run the devices command:

    

        
        
            adb devices
        
    

If the Android device is found it will be listed with an ID number. The output will be similar to:

    

        
        List of devices attached
00000027A00013  device

        
    

If no devices are listed, there is something wrong with the connection between your computer and the Android device.

4. Copy the application to the Android device:

You are now ready to copy the application to the Android device and run it.

Use adb push to copy the file to the device and give it execute permission:

    

        
        
            adb push ./bin/arm_gpuinfo /data/local/tmp
adb shell chmod u+x /data/local/tmp/arm_gpuinfo
        
    

Run the application:

    

        
        
            adb shell /data/local/tmp/arm_gpuinfo
        
    

The output will be similar to:

    

        
        Device configuration
  - Manufacturer: Khadas
  - Model: Edge2
  - Android version: 12
  - Kernel version: 5.10.110

GPU configuration
  - Name: Mali-G610
  - Architecture: Valhall
  - Model number: 0xa007
  - Core count: 4
  - L2 cache count: 4
  - Total L2 cache size: 1048576 bytes
  - Bus width: 128 bits

Per-core statistics
  - Engine count: 2
  - FP32 FMAs: 64/cy
  - FP16 FMAs: 128/cy
  - Texels: 8/cy
  - Pixels: 4/cy

Per-GPU statistics
  - FP32 FMAs: 256/cy
  - FP16 FMAs: 512/cy
  - Texels: 32/cy
  - Pixels: 16/cy

        
    

The configuration details will vary based on the Arm GPU and Android device you are using. The output above is from the Khadas Edge 2 single-board computer running Android 12. It includes a 4-core Mali-G610 GPU.

You have learned how to build and run a program to retrieve the hardware configuration details for Arm Mali and Immortalis GPUs.