How to Build M68k LLVM

For users of M68k LLVM

Building M68k LLVM is roughly the same as building a normal LLVM.

Perequisites:

• A Unix-like OS
• A git client
• A native C++ compiler
• CMake
• Ninja build system

CMake usage can be displayed with cmake --help.

We’ll use a directory named build to store build files. This directory is already ignored by the .gitignore file, so it’s assumed to be the standard build directory.

$ git clone https://github.com/llvm/llvm-project.git
$ cd llvm-project
$ mkdir build
$ cd build
$ cmake -G Ninja -D CMAKE_BUILD_TYPE="Release" -D LLVM_EXPERIMENTAL_TARGETS_TO_BUILD="M68k" -S ../llvm
$ cmake --build .

The LLVM CMake options which affect the build are documented here.

The main difference from a typical LLVM configuration is the LLVM_EXPERIMENTAL_TARGETS_TO_BUILD CMake argument, since M68k is not an official target yet.

-G Ninja

We’ll use the Ninja build system, rather than the default (Make on Unix-like OSes.) Ninja is faster than Make.

CMAKE_BUILD_TYPE tells CMake if we want optimisations, debug symbols or LLVM assertions. Release enables optimisations, and disables both debug symbols and LLVM assertions.

-S ../llvm

Specifies ../llvm as the source directory.

If you want to build Clang, make sure to add the -DLLVM_ENABLE_PROJECTS="clang" cmake argument as well.

For more details about the CMake build process and various options, please refer to the official documentation.

For contributors to M68k LLVM

Example 1: Configuring for general development

The following cmake configurations may be used for a development setup on Linux:

cmake \
    -D BUILD_SHARED_LIBS="On" \
    -D CMAKE_BUILD_TYPE="Debug" \
    -D LLVM_ENABLE_PROJECTS="clang;compiler-rt" \
    -D LLVM_EXPERIMENTAL_TARGETS_TO_BUILD="M68k" \
    -D LLVM_TARGETS_TO_BUILD="Native" \
    -D LLVM_USE_LINKER="lld" \
    -D LLVM_USE_SPLIT_DWARF="On" \
    -G Ninja \
    -S ../llvm

-D BUILD_SHARED_LIBS="On"

We’ll build shared libraries rather than static libraries, to save some disk space.

You need to install LLD for this configuration (or you can replace lld in the above command with gold), which is available in most of the Debian / Ubuntu distributions.

-D LLVM_USE_SPLIT_DWARF="On"

This option can reduce memory usage at link-time.

Example 2: Configuring for running of LLVM tests

cmake \
    -D BUILD_SHARED_LIBS="On" \
    -D CMAKE_BUILD_TYPE="RelWithDebInfo" \
    -D LLVM_ENABLE_ASSERTIONS="On" \
    -D LLVM_ENABLE_PROJECTS="" \
    -D LLVM_EXPERIMENTAL_TARGETS_TO_BUILD="M68k" \
    -D LLVM_TARGETS_TO_BUILD="" \
    -D LLVM_USE_SPLIT_DWARF="On" \
    -G Ninja \
    -S ../llvm

-D CMAKE_BUILD_TYPE="RelWithDebInfo"

We’ll build the RelWithDebInfo configuration rather than Debug, as otherwise running tests can be quite slow.

We could alternatively build the Release or MinSizeRel configuration instead, but debug information can be useful.

-D LLVM_ENABLE_ASSERTIONS="On"

We’ll enable LLVM assertions, in order to catch more errors during runtime. Specifying a configuration other than Debug normally disables these assertions.

-D LLVM_ENABLE_PROJECTS=""

We can build more quickly by removing unnecessary projects from the build. If you’re working with Clang, specify clang.

-D LLVM_TARGETS_TO_BUILD=""

We’ll remove all other targets from the build. This can speed up the build if we don’t need the native LLVM target.