dfg-mlir

This repository implements an MLIR dialect for representing dataflow graphs.

Building

The dfg-mlir project is built using CMake (version 3.20 or newer). Make sure to provide all dependencies required by the project, either by installing them to system-default locations, or by setting the appropriate search location hints!

Using nix

To build the dialect on a Nix-based system, you can use the provided flake.nix file to get a development shell up and running. Make sure that you have enabled Flakes for your Nix installation (verifiable by calling nix flake from the command line).

Then, just run nix develop to get a shell that provides LLVM and MLIR versions working for this project. Build the dialect by running

# Configure.
cmake -S . -B build $cmakeFlags
# Build.
cmake --build build

Manually

Make sure, you have MLIR and LLVM built and available on your system. Then point CMake to the include directories as shown below:

# Configure.
cmake -S . -B build \
    -G Ninja \
    -DLLVM_DIR=$LLVM_PREFIX/lib/cmake/llvm \
    -DMLIR_DIR=$MLIR_PREFIX/lib/cmake/mlir \
    -DCIRCT_DIR=$CIRCT_PREFIX/lib/cmake/circt
    (Optional)
    -DCMAKE_C_COMPILER=<clang> \
    -DCMAKE_CXX_COMPILER=<clang++> \
    -DLLVM_USE_LINKER=<lld>

# Build.
cmake --build build

The following CMake variables can be configured:

Name	Type	Description
LLVM_DIR	STRING	Path to the CMake directory of an LLVM installation. e.g. ~/tools/llvm-17/lib/cmake/llvm
MLIR_DIR	STRING	Path to the CMake directory of an MLIR installation. e.g. ~/tools/llvm-17/lib/cmake/mlir
CIRCT_DIR	STRING	Path to the CMake directory of an CIRCT installation. e.g. ~/tools/circt/lib/cmake/circt

Notice that the llvm version should be the same as the CIRCT is using.

Acknowledgements

dfg-mlir has been supported throughout its history by the following projects.

European Union projects:

• Grant agreement ID 957269 EVEREST – dEsign enVironmEnt foR Extreme-Scale big data analytics on heterogeneous platforms