MLIR-Based Ideas Landing Project (Project page).

This project uses LLVM/MLIR as an external library. Please make sure the dependencies are available on your machine.

$ git clone [email protected]:buddy-compiler/buddy-mlir.git
$ cd buddy-mlir
$ git submodule update --init

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

If your target machine includes a Nvidia GPU, you can use the following configuration:

$ cmake -G Ninja ../llvm \
    -DLLVM_ENABLE_PROJECTS="mlir;clang" \
    -DLLVM_TARGETS_TO_BUILD="host;RISCV;NVPTX" \
    -DMLIR_ENABLE_CUDA_RUNNER=ON \
    -DLLVM_ENABLE_ASSERTIONS=ON \
    -DCMAKE_BUILD_TYPE=RELEASE

If your target machine has lld installed, you can use the following configuration:

$ cmake -G Ninja ../llvm \
    -DLLVM_ENABLE_PROJECTS="mlir;clang" \
    -DLLVM_TARGETS_TO_BUILD="host;RISCV" \
    -DLLVM_USE_LINKER=lld \
    -DLLVM_ENABLE_ASSERTIONS=ON \
    -DCMAKE_BUILD_TYPE=RELEASE

$ cd buddy-mlir
$ mkdir build
$ cd build
$ cmake -G Ninja .. \
    -DMLIR_DIR=$PWD/../llvm/build/lib/cmake/mlir \
    -DLLVM_DIR=$PWD/../llvm/build/lib/cmake/llvm \
    -DLLVM_ENABLE_ASSERTIONS=ON \
    -DCMAKE_BUILD_TYPE=RELEASE
$ ninja
$ ninja check-buddy

If you want to add domain-specific framework support, please add the following cmake options:

Bud dialect is designed for testing and demonstrating.

DIP dialect is designed for digital image processing abstraction.

The buddy-opt is the driver for dialects and optimization in buddy-mlir project.

The AutoConfig mechanism is designed to detect the target hardware and configure the toolchain automatically.

The purpose of the examples is to give users a better understanding of how to use the passes and the interfaces in buddy-mlir. Currently, we provide three types of examples.

• IR level conversion and transformation examples.
• Domain-specific application level examples.
• Testing and demonstrating examples.

For more details, please see the documentation of the examples.

The benchmarks in this repo use JIT tool (mlir-cpu-runner) as the execution engine. For AOT benchmarks, please see buddy-benchmark repo.

We provide the following benchmarks:

• Conv2D

$ cd buddy-mlir/benchmark
$ make

For more features and configurations, please see the benchmark document.