c-mlir

Prerequisites

• C/C++ toolchains (compiler, linker etc.)
• cmake
• make or ninja.
• lit(optional for testing)

Installation

Option 1: Build from scratch

1. Configuration:

make config # generate configuration file at build/config.cmake

2. Build LLVM, MLIR and Clang:

mkdir llvm-project/build
cd llvm-project/build
cmake -G Ninja ../llvm \
  -DLLVM_ENABLE_PROJECTS="mlir;clang" \
  -DLLVM_TARGETS_TO_BUILD="host" \
  -DLLVM_ENABLE_ASSERTIONS=ON \
  -DCMAKE_BUILD_TYPE=DEBUG
ninja
ninja check-mlir

For faster compilation,it is recommended to use -DLLVM_USE_LINKER=lld.

3. Build cmlirc:

make

Or manually follow this:

mkdir build
cd build
cmake -G Ninja ..
ninja

Option 2: Build with pre-built LLVM, MLIR and Clang

1. Configuration:

make config # generate configuration file at build/config.cmake

Specify the PATH in config.cmake.

# Build configuration options
# NOTE: set(LLVM_BUILD_DIR path/to/your/pre-built/installation) # if you're using your pre-built installation.
set(LLVM_BUILD_DIR ${CMAKE_SOURCE_DIR}/llvm-project/build) # If you're building this proj from scratch, you should not change it

# Compilation options
set(USE_CCACHE ON)

# Build options
set(BUILD_EXECUTABLES ON)
set(ENABLE_TESTING ON) # Set it ON to enable testing

# Derived paths
# NOTE: you can force cmake to overwrite these PATHs
set(LLVM_DIR ${LLVM_BUILD_DIR}/lib/cmake/llvm)
set(MLIR_DIR ${LLVM_BUILD_DIR}/lib/cmake/mlir)
set(Clang_DIR ${LLVM_BUILD_DIR}/lib/cmake/clang)

2. Build cmlirc:

make

Or manually follow this:

mkdir build
cd build
cmake -G Ninja ..
ninja

Run Tests

1. Install lit

Activate a python virtual env(venv, conda, uv etc.) and install lit using pip:

pip install lit

2. Run tests

make test

Or manually follow this:

cd build
ninja check-cmlirc

Acknowledgement

• This work is inspired by Polygeist

License

Copyright 2026 dashygo097

Licensed under the Apache License, Version 2.0. See LICENSE for details.