/circt

Circuit IR Compilers and Tools

Primary LanguageC++OtherNOASSERTION

⚡️ "CIRCT" / Circuit IR Compilers and Tools

"CIRCT" stands for "Circuit IR Compilers and Tools". One might also interpret it as the recursively as "CIRCT IR Compiler and Tools". The T can be selectively expanded as Tool, Translator, Team, Technology, Target, Tree, Type, ... we're ok with the ambiguity.

The CIRCT community is an open and welcoming community. If you'd like to participate, you can do so in a number of different ways:

  1. Join our Discourse Forum on the LLVM Discourse server. To get a "mailing list" like experience click the bell icon in the upper right and switch to "Watching". It is also helpful to go to your Discourse profile, then the "emails" tab, and check "Enable mailing list mode".

  2. Join our weekly video chat. Please see the meeting notes document for more information.

  3. Contribute code. CIRCT follows all of the LLVM Policies: you can create pull requests for the CIRCT repository, and gain commit access using the standard LLVM policies.

Motivation

The EDA industry has well-known and widely used proprietary and open source tools. However, these tools are inconsistent, have usability concerns, and were not designed together into a common platform. Furthermore these tools are generally built with Verilog (also VHDL) as the IRs that they interchange. Verilog has well known design issues, and limitations, e.g. suffering from poor location tracking support.

The CIRCT project is an (experimental!) effort looking to apply MLIR and the LLVM development methodology to the domain of hardware design tools. Many of us dream of having reusable infrastructure that is modular, uses library-based design techniques, is more consistent, and builds on the best practices in compiler infrastructure and compiler design techniques.

By working together, we hope that we can build a new center of gravity to draw contributions from the small (but enthusiastic!) community of people who work on open hardware tooling. In turn we hope this will propel open tools forward, enables new higher-level abstractions for hardware design, and perhaps some pieces may even be adopted by proprietary tools in time.

For more information, please see our longer charter document.

Setting this up

These commands can be used to set this up on a Mac:

  1. Install Dependencies of LLVM/MLIR according to the instructions, including cmake and ninja.

  2. Check out LLVM and CIRCT repos. CIRCT contains LLVM as a git submodule. The LLVM repo here includes staged changes to MLIR which may be necessary to support CIRCT. It also represents the version of LLVM that has been tested. MLIR is still changing relatively rapidly, so feel free to use the current version of LLVM, but APIs may have changed.

$ git clone git@github.com:circt/circt.git
$ cd circt
$ git submodule init
$ git submodule update
  1. Build and test LLVM/MLIR:
$ cd circt
$ mkdir llvm/build
$ cd llvm/build
$ cmake -G Ninja ../llvm \
    -DLLVM_ENABLE_PROJECTS="mlir" \
    -DLLVM_TARGETS_TO_BUILD="X86;RISCV" \
    -DLLVM_ENABLE_ASSERTIONS=ON \
    -DCMAKE_BUILD_TYPE=DEBUG
$ ninja
$ ninja check-mlir
  1. Build and test CIRCT:
$ cd circt
$ 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=DEBUG
$ ninja
$ ninja check-circt

The -DCMAKE_BUILD_TYPE=DEBUG flag enables debug information, which makes the whole tree compile slower, but allows you to step through code into the LLVM and MLIR frameworks.

To get something that runs fast, use -DCMAKE_BUILD_TYPE=Release or -DCMAKE_BUILD_TYPE=RelWithDebInfo if you want to go fast and optionally if you want debug info to go with it. Release mode makes a very large difference in performance.

  1. Optionally configure your environment:

It is useful to add the .../circt/build/bin and .../circt/llvm/build/bin directories to the end of your PATH, allowing you to use the tools like circt-opt in a natural way on the command line. Similarly, you need to be in the build directory to invoke ninja, which is super annoying. You might find a bash/zsh alias like this to be useful:

build() {
  (cd $HOME/Projects/circt/build/; ninja $1 $2 $3)
}

This allows you to invoke build check-circt from any directory and have it do the right thing.

Submitting changes to CIRCT

The project is small so there are few formal process yet. We generally follow the LLVM and MLIR community practices, but we currently use pull requests and GitHub issues. Here are some high-level guidelines:

  • Please use clang-format in the LLVM style. There are good plugins for common editors like VSCode, Atom, etc, or you can run it manually. This makes code easier to read and understand.

  • Beyond mechanical formatting issues, please follow the LLVM Coding Standards.

  • Please practice "incremental development", preferring to send a small series of incremental patches rather than large patches. There are other policies in the LLVM Developer Policy document that are worth skimming.

  • Please use "Squash and Merge" in PRs when they are approved - we don't need the intra-change history in the repository history.

  • Please create a PRs to get a code review. For reviewers, it is good to look at the primary author of the code you are touching to make sure they are at least CC'd on the PR.

Submitting changes to LLVM / MLIR

This project depends on MLIR and LLVM, and it is occasionally useful to improve them. To get set up for this:

  1. Follow the "How to Contribute" instructions, and install the right tools, e.g. 'arcanist' and clang-format.
  2. Get an LLVM Phabricator account
  3. Ask for LLVM commit access, the barrier is low.

Submitting a patch

The patch flow goes like this:

  1. Check out the LLVM mono repo (as described above).
  2. Make changes to your codebase.
  3. Stage your changes with git add.
  4. Run clang-format to tidy up the details of the patch with git clang-format origin/master
  5. Create a patch in Phabricator with arc diff.
  6. Iterate on review, changing your code and sending another patch with arc diff.

When your review converges and your patch is approved, do the following:

  1. commit your changes with git commit
  2. rebase your changes with git pull --rebase
  3. retest your patch with ninja check-mlir (or whatever other target makes sense)
  4. push your changes with git push

Updating the staging version of LLVM / MLIR

The 'staging' tree at git@github.com:circt/llvm.git is only intended to be a temporary location for patches which have not yet been accepted upstream. The patches here should not be long lived and the branches in that tree may be rebased without warning. The 'correct' version of LLVM is referenced by its git hash as a submodule of the main circt repository. It may be necessary to pull changes from LLVM, in which case it is appropriate to rebase this tree.

$ cd llvm
$ git remote add llvm git@github.com:llvm/llvm-project.git
$ git pull --rebase llvm master
$ cd build
$ ninja
$ ninja check-mlir

At this point we have built and tested our changes with any upstream changes and resolved any conflicts. Note that we rebase instead of merging in order to ensure that our patches remain current, since they should get pushed to LLVM at some point. Now we can update circt/llvm and push a new reference to LLVM for everyone else to use.

$ git push -f origin HEAD:staging

However, we're still not done at this point. circt/circt still points at the old git hash for the llvm submodule.

$ cd circt/build
$ ninja
$ ninja check-circt

Now we have fixed any changes to circuit, so those changes can be committed and pushed.

$ cd ..
$ git commit .
$ git push origin HEAD:master

You should see that the changes include an update to the submodule hash.