and you'll miss it...

To build the frontend of the language (parser, typechecker, desugarer) we need a few OCaml libraries.

Before we install any of them, first I recommend creating a custom opam switch to compile our OCaml code using Clang rather than GCC. I have a short Github gist here explaining how to do this.

Next, run opam install . --deps-only in frontend/ which will install all the necessary dependencies listed in blink.opam. I've seen issues where ounit2 won't be installed with this, but installing it separately should work.

To build the backend of the langauge (LLVM codegen) we need to install LLVM. This project currently uses LLVM 16.0.0; I have plans to update to a newer version soon. I specifically used LLVM 16.0.0 installed to my machine from the Github repo.

Here are some steps to install LLVM:

sudo apt install -y cmake ninja-build build-essential python3-dev libz-dev libxml2-dev

git clone --depth 1 --branch llvmorg-16.0.0 https://github.com/llvm/llvm-project.git

cd llvm-project

mkdir build && cd build

cmake -G Ninja -S ../llvm -B . \
  -DCMAKE_BUILD_TYPE=Release \
  -DLLVM_ENABLE_PROJECTS="clang;lld;clang-tools-extra" \
  -DLLVM_TARGETS_TO_BUILD="AArch64;AMDGPU;ARM;NVPTX;X86;XCore;RISCV" \
  -DCMAKE_CXX_STANDARD=17 \
  -DLLVM_ENABLE_ASSERTIONS=OFF \
  -DCMAKE_INSTALL_PREFIX=/usr/local/llvm16

ninja  # might need to specify parallel jobs with -j <cores>

sudo ninja install

Running make in the root directory of this project builds the entire compiler. This produces the blink executable. Running this on a .bl file will produce an LLVM-IR file called new_output.ll.

To build the frontend, you can simply run dune build in frontend/.

Take a look at the examples in examples/. You can compile a program to an executable using ./compile which will generate new_output.o. Take a look at the compile script if you want to customize the final executable.