Shortcuts

Building with CMake

ExecuTorch uses CMake as its primary build system. Even if you don’t use CMake directly, CMake can emit scripts for other format like Make, Ninja or Xcode. For information, see cmake-generators(7).

Targets Built by the CMake Build System

ExecuTorch’s CMake build system covers the pieces of the runtime that are likely to be useful to embedded systems users.

  • libexecutorch.a: The core of the ExecuTorch runtime. Does not contain any operator/kernel definitions or backend definitions.

  • libportable_kernels.a: The implementations of ATen-compatible operators, following the signatures in //kernels/portable/functions.yaml.

  • libportable_kernels_bindings.a: Generated code that registers the contents of libportable_kernels.a with the runtime.

    • NOTE: This must be linked into your application with a flag like -Wl,-force_load or -Wl,--whole-archive. It contains load-time functions that automatically register the kernels, but linkers will often prune those functions by default because there are no direct calls to them.

  • executor_runner: An example tool that runs a .pte program file using all 1 values as inputs, and prints the outputs to stdout. It is linked with libportable_kernels.a, so the program may use any of the operators it implements.

One-time setup to prepare for CMake Build

Follow the steps below to have the tools ready before using CMake to build on your machine.

  1. If your system’s version of python3 is older than 3.11:

    • Run pip install tomli

  2. Install CMake version 3.19 or later:

    • Run conda install cmake or pip install cmake.

Configure the CMake Build

Follow these steps after cloning or pulling the upstream repo, since the build dependencies may have changed.

# cd to the root of the executorch repo
cd executorch

# Clean and configure the CMake build system. It's good practice to do this
# whenever cloning or pulling the upstream repo.
(rm -rf cmake-out && mkdir cmake-out && cd cmake-out && cmake ..)

Once this is done, you don’t need to do it again until you pull from the upstream repo again, or if you modify any CMake-related files.

CMake Build Options

The release build offers optimizations intended to improve performance and reduce binary size. It disables program verification and executorch logging, and adds optimizations flags.

-DCMAKE_BUILD_TYPE=Release

To further optimize the release build for size, use both:

-DCMAKE_BUILD_TYPE=Release \
-DOPTIMIZE_SIZE=ON

See CMakeLists.txt

Build the runtime components

Build all targets with

# cd to the root of the executorch repo
cd executorch

# Build using the configuration that you previously generated under the
# `cmake-out` directory.
#
# NOTE: The `-j` argument specifies how many jobs/processes to use when
# building, and tends to speed up the build significantly. It's typical to use
# "core count + 1" as the `-j` value.
cmake --build cmake-out -j9

Use an example app executor_runner to execute a .pte file

First, generate an add.pte or other ExecuTorch program file using the instructions as described in Setting up ExecuTorch.

Then, pass it to the command line tool:

./cmake-out/executor_runner --model_path path/to/add.pte

If it worked, you should see the message “Model executed successfully” followed by the output values.

I 00:00:00.002052 executorch:executor_runner.cpp:75] Model file add.pte is loaded.
I 00:00:00.002086 executorch:executor_runner.cpp:85] Running method forward
I 00:00:00.002092 executorch:executor_runner.cpp:140] Setting up non-const buffer 1, size 48.
I 00:00:00.002149 executorch:executor_runner.cpp:181] Method loaded.
I 00:00:00.002154 executorch:util.h:105] input already initialized, refilling.
I 00:00:00.002157 executorch:util.h:105] input already initialized, refilling.
I 00:00:00.002159 executorch:executor_runner.cpp:186] Inputs prepared.
I 00:00:00.011684 executorch:executor_runner.cpp:195] Model executed successfully.
I 00:00:00.011709 executorch:executor_runner.cpp:210] 8.000000

Cross compilation

Follwing are instruction on how to perform cross compilation for Android and iOS.

Android

  • Prerequisite: Android NDK, choose one of the following:

    • Option 1: Download Android Studio by following the instructions to install ndk.

    • Option 2: Download Android NDK directly from here.

Assuming Android NDK is available, run:

# Run the following lines from the `executorch/` folder
rm -rf cmake-android-out && mkdir cmake-android-out && cd cmake-android-out

# point -DCMAKE_TOOLCHAIN_FILE to the location where ndk is installed
cmake -DCMAKE_TOOLCHAIN_FILE=/Users/{user_name}/Library/Android/sdk/ndk/25.2.9519653/build/cmake/android.toolchain.cmake  -DANDROID_ABI=arm64-v8a ..

cd  ..
cmake --build  cmake-android-out  -j9

adb shell mkdir -p /data/local/tmp/executorch
# push the binary to an Android device
adb push  cmake-android-out/executor_runner  /data/local/tmp/executorch
# push the model file
adb push  add.pte  /data/local/tmp/executorch

adb shell  "/data/local/tmp/executorch/executor_runner --model_path /data/local/tmp/executorch/add.pte"

iOS

For iOS we’ll build frameworks instead of static libraries, that will also contain the public headers inside.

  1. Install Xcode from the Mac App Store and then install the Command Line Tools using the terminal:

xcode-select --install
  1. Build the frameworks:

./build/build_apple_frameworks.sh

Run the above command with --help flag to learn more on how to build additional backends (like Core ML, MPS or XNNPACK), etc. Note, some backends may require additional dependencies and certain versions of Xcode and iOS.

  1. Copy over the generated .xcframework bundles to your Xcode project, link them against your targets and don’t forget to add an extra linker flag -all_load.

Check out the iOS Demo App tutorial for more info.

Next steps

You have successfully cross-compiled executor_runner binary to iOS and Android platforms. You can start exploring advanced features and capabilities. Here is a list of sections you might want to read next:

  • Selective build to build the runtime that links to only kernels used by the program, which can provide significant binary size savings.

  • Tutorials on building Android and iOS demo apps.

  • Tutorials on deploying applications to embedded devices such as ARM Cortex-M/Ethos-U and XTensa HiFi DSP.

Docs

Access comprehensive developer documentation for PyTorch

View Docs

Tutorials

Get in-depth tutorials for beginners and advanced developers

View Tutorials

Resources

Find development resources and get your questions answered

View Resources