Skip to content

build.md

Latest commit

 

History

History
109 lines (85 loc) · 5.04 KB

build.md

File metadata and controls

109 lines (85 loc) · 5.04 KB

Building the code

Overview

The C++ library clij2fft implements several FFT-based algorithms built on top of clFFT. This library is independent of Java.

Note: For Java, native libraries are now included in the JAR, so there is no need to manually copy them to an update site or Fiji installation.

Building Native Libraries

Using GitHub Actions (Recommended)

  • No local native builds are required or recommended.
  • If you modify the native code, push your changes to GitHub with the commit message containing [build natives].
  • GitHub Actions will build the native libraries for all platforms and commit the updated libraries back to the repo.
  • Pull the latest changes to get the updated native libraries.

Java release cycle

Building the Java Plugin

If you modified the native, code, build the new natives following the new workflow. Then run the following Maven command to build the jar:

mvn clean package "-Dgpg.skip=true" "-Djavacpp.skip=true"

Installation of the Java plugin in your local Fiji

Once the Java plugin is built, copy the JAR (./target/clij2-fft-{version}.jar) to your Fiji installation (Fiji.app/plugins/).

Java releases to scijava maven

To create a release:

  1. Ensure all native libraries are up-to-date (built via GitHub Actions).
  2. Use the release-version.sh script from scijava-scripts with the following arguments to skip native builds and GPG signing: 
    ./release-version.sh -Darguments="-Dgpg.skip=true -Djavacpp.skip=true"

Troubleshooting

  • Search the Image.sc Forum for more information.
  • Ask questions on the forum if previous discussions are unclear.
Building Native Libraries Locally - Legacy (Click to expand) ## Building Native Libraries - Legacy

Old Workflow (Local Builds)

If you still need to build natives locally (e.g., for testing or development), follow the instructions below.

Prerequisites for All Operating Systems

Windows

Prerequisites
Build Instructions
  1. From the Start menu, run 'x64 Native Tools Command Prompt for VS'.
  2. From the Command Prompt, start a bash shell: "C:\Program Files\Git\bin\sh.exe".
  3. Run ./native/cppbuild.sh.
  4. If step 3 fails, check ./native/clij2fft/cppbuild.sh and verify that OpenCL and clFFT are installed in the correct locations.
  5. The updated library (clij2fft.dll) and dependencies will be in the ./lib/windows-x86_64/ directory.

Linux

Prerequisites
Build Instructions
  1. From a bash terminal, run ./native/cppbuild.sh.
  2. If step 1 fails, check ./native/clij2fft/cppbuild.sh and verify that OpenCL and clFFT are installed in the correct locations.
  3. The updated library (clij2fft.so) and dependencies will be in the ./lib/linux-x86_64/ directory.

macOS (Intel and Apple Silicon M1/M2/M3)

Prerequisites
  • clFFT needs to be built from source 
     brew install cmake || true
 git clone https://github.com/clMathLibraries/clFFT.git
 cd clFFT/src
 sed -i.bak 's/cmake_minimum_required( VERSION 3.1 )/cmake_minimum_required( VERSION 3.6 )/' CMakeLists.txt
 mkdir build && cd build
 cmake .. -DCMAKE_BUILD_TYPE=Release -DBUILD_SHARED_LIBS=ON -DCMAKE_INSTALL_PREFIX=${{ matrix.lib_prefix }}
 make -j$(sysctl -n hw.ncpu) && sudo make install
Build Instructions
  1. From a bash terminal, run ./native/cppbuild.sh.
  2. If step 1 fails, check ./native/clij2fft/cppbuild.sh and verify that OpenCL and clFFT are installed in the correct locations.
  3. For macOS, the native library libclij2fft needs to be modified with install_name_tool to find clFFT in its current directory: 
    install_name_tool -change libclFFT.2.dylib @rpath/libclFFT.2.dylib ../../../lib/macosx-arm64/libclij2fft.dylib
  4. If targeting both Intel and Apple Silicon, you can build a universal binary using lipo: 
    lipo -create -output lib/macosx-universal2/libclFFT.dylib lib/macosx-x86_64/libclFFT.dylib lib/macosx-arm64/libclFFT.dylib
  5. The updated library (clij2fft.dylib) and dependencies will be in the ./lib/macosx-x86_64/ or ./lib/macosx-arm64/ directory.