Extended Installation (ACL)

ACL acceleration optimizes the performance of tasks such as deep learning in the Arm architecture. Replace the example paths provided in the following procedure with the actual ones.

Obtain the ACL source code.
1. Download the ACL source code.
  https://github.com/ARM-software/ComputeLibrary/archive/refs/tags/v23.08.tar.gz
2. Use an SFTP tool to upload the source package to the server.
If the server is connected to the Internet, you can run the wget command on the server to download the source package.
```
wget https://github.com/ARM-software/ComputeLibrary/archive/refs/tags/v23.08.tar.gz --no-check-certificate
```
Decompress the ACL source package.
```
tar -xzvf v23.08.tar.gz
```
Go to the /path/to/ComputeLibrary-23.08 directory.
```
cd /path/to/ComputeLibrary-23.08
```

Build scons.

scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=linux arch=armv8a build=native

Set the environment variable.

export ACL_ROOT_DIR=/path/to/ComputeLibrary-23.08

Verify the installation.

Go to the build directory.
```
cd build
```

Set the environment variable.

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/ComputeLibrary-23.08/build/

Run the example test case.
```
./examples/neon_cnn
```

Install oneDNN following the instructions in Basic Installation. Replace 7 with the following command to complete the oneDNN compilation and installation.
```
cmake -DCMAKE_INSTALL_PREFIX=/path/to/oneDNN-3.3.3/build/install/ -DONEDNN_AARCH64_USE_ACL=ON ..
```
After the installation is complete, run the following command to check whether libdnnl.so is linked to the SO file of the ACL library:
```
ldd /path/to/oneDNN-3.3.3/build/install/lib64/libdnnl.so
```
If the following information is displayed, libdnnl.so has been linked to the SO file of the ACL library.

oneDNN 3.3.3 conflicts with ACL 23.11 and later. If a compilation error occurs, check whether the software version is the correct version. If oneDNN 3.3.3 is used, the ACL version must be 23.08 or earlier.

Parent topic: Compiling the Source Code