Computational physics on the smartphone GPU

Screenshot of the interacting many-body simulation on the Nexus 4 GPU.
Screenshot of the interacting many-body simulation on the Nexus 4 GPU.

[Update August 2013: Google has removed the OpenCL library with Android 4.3. You can find an interesting discussion here. Google seems to push for its own renderscript protocol. I will not work with renderscript since my priorities are platform independency and sticking with widely adopted  standards to avoid fragmentation of my code basis.]
I recently got hold of a Nexus 4 smartphone, which features a GPU (Qualcomm Adreno 320) and conveniently ships with already installed OpenCL library. With minimal changes I got the previously discussed many-body program code related to the fractional quantum Hall effect up and running. No unrooting of the phone is required to run the code example. Please use the following recipe at your own risk, I don’t accept any liabilities. Here is what I did:

  1. Download and unpack the Android SDK from google for cross-compilation (my host computer runs Mac OS X).
  2. Download and unpack the Android NDK from google to build minimal C/C++ programs without Java (no real app).
  3. Install the standalone toolchain from the Android NDK. I used the following command for my installation:

    /home/tkramer/android-ndk-r8d/build/tools/ \
  4. Put the OpenCL programs and source code in an extra directory, as described in my previous post
  5. Change one line in the cl.hpp header: instead of including <GL/gl.h> change to <GLES/gl.h>. Note: I am using the “old” cl.hpp bindings 1.1, further changes might be required for the newer bindings, see for instance this helpful blog
  6. Transfer the OpenCL library from the phone to a subdirectory lib/ inside your source code. To do so append the path to your SDK tools and use the adb command:

    export PATH=/home/tkramer/adt-bundle-mac-x86_64-20130219/sdk/platform-tools:$PATH
    adb pull /system/lib/
  7. Cross compile your program. I used the following script, please feel free to provide shorter versions. Adjust the include directories and library directories for your installation.

    rm plasma_disk_gpu
    /home/tkramer/android-ndk-standalone/bin/arm-linux-androideabi-g++ -v -g \
    -I. \
    -I/home/tkramer/android-ndk-standalone/include/c++/4.6 \
    -I/home/tkramer/android-ndk-r8d/platforms/android-5/arch-arm/usr/include \
    -Llib \
    -march=armv7-a -mfloat-abi=softfp -mfpu=neon \
    -fpic -fsigned-char -fdata-sections -funwind-tables -fstack-protector \
    -ffunction-sections -fdiagnostics-show-option -fPIC \
    -fno-strict-aliasing -fno-omit-frame-pointer -fno-rtti \
    -lOpenCL \
    -o plasma_disk_gpu plasma_disk.cpp
  8. Copy the executable to the data dir of your phone to be able to run it. This can be done without rooting the phone with the nice SSHDroid App, which by defaults transfers to /data . Don’t forget to copy the kernel .cl files:

    scp -P 2222 root@192.168.0.NNN:
    scp -P 2222 plasma_disk_gpu root@192.168.0.NNN:
  9. ssh into your phone and run the GPU program:
    ssh -p 2222 root@192.168.0.NNN
    ./plasma_disk_gpu 64 16
  10. Check the resulting data files. You can copy them for example to the Download path of the storage and use the gnuplot (droidplot App) to plot them.

A short note about runtimes. On the Nexus 4 device the program runs for about 12 seconds, on a MacBook Pro with NVIDIA GT650M it completes in 2 seconds (in the example above the equations of motion for 16*64=1024 interacting particles are integrated). For larger particle numbers the phone often locks up.

An alternative way to transfer files to the device is to connect via USB cable and to install the Android Terminal Emulator app. Next

cd /data/data/jackpal.androidterm
mkdir gpu
chmod 777 gpu

On the host computer use adb to transfer the compiled program and the .cl kernel and start a shell to run the kernel

adb push /data/data/jackpal.androidterm/gpu/
adb push plasma_disk_gpu /data/data/jackpal.androidterm/gpu/

You can either run the program within the terminal emulator or use the adb shell

adb shell
cd /data/data/jackpal.androidterm/gpu/
./plasma_disk_gpu 64 16

Let’s see in how many years todays desktop GPUs can be found in smartphones and which computational physics codes can be run!

2 thoughts on “Computational physics on the smartphone GPU

  1. Amazing stuff, OpenCL will have a large impact on the mobile platforms in the coming years. Well done!

  2. After a long stagnation OpenCL on mobile phones is resurfacing again:

    Intel included in their 2014 OpenCL release Android support (for intel chips), and some sites posted ways to enable OpenCL on a Nexus 5 device running Android 4.4.2
    ( look for instance on ).
    I can confirm this works for my Nexus 4 mentioned in the blog post, but is not officially supported and requires rooting your phone, so proceed with caution and at your own risk.
    The plus side: these drivers work way more stable than the previous ones distributed by google and finally one can run longer-time and larger memory kernels without crashing the phone. Additionally Sony seems to include OpenCL on some phones and Mali also supports OpenCL on their chips.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.