Awesome

Firefly - A new compiler and runtime for BEAM languages

Machine	Vendor	Operating System	Host	Subgroup
wasm32	unknown	unknown	macOS	N/A
wasm32	unknown	unknown	Linux	N/A
x86_64	apple	darwin	macOS	compiler
x86_64	apple	darwin	macOS	libraries
x86_64	apple	darwin	macOS	firefly/otp
x86_64	apple	darwin	macOS	runtime full
x86_64	unknown	linux-gnu	Linux	libraries
x86_64	unknown	linux-gnu	Linux	firefly/otp
x86_64	unknown	linux-gnu	Linux	runtime full

Contributing
About Firefly
Goals
Non-Goals
Architecture

Contributing

In order to build Firefly, or make changes to it, you'll need the following installed:

Tools

First, you will need to install rustup. Follow the instructions at that link.

Once you have installed rustup, you will need to install the nightly version of Rust (currently our CI builds against the 2022-07-12 nightly, specifically). We require nightly due to a large number of nightly features we use, as well as some dependencies for the WebAssembly targets that we make use of.

# to use the latest nightly
rustup default nightly

# or, in case of issues, install the 2022-07-12 nightly to match our CI
rustup default nightly-2022-07-12
export CARGO_MAKE_TOOLCHAIN=nightly-2022-07-12

In order to run various build tasks in the project, you'll need the cargo-make plugin for Cargo. You can install it with:

cargo install cargo-make

You can see what tasks are available with cargo make --print-steps.

You may also want to install the following tools for editor support (rustfmt will be required on all pull requests!):

rustup component add rustfmt clippy

Next, for wasm32 support, you will need to install the wasm32 targets for the Rust toolchain:

rustup target add wasm32-unknown-unknown --toolchain <name of nightly you chose in the previous step>

LLVM

LLVM (with some patches of our own) is used internally for the final code generation stage. In order to build the compiler, you must have our LLVM installed somewhere locally. Typically, you'd need to build this yourself, which we have instructions for below; but we also produce prebuilt packages that have everything needed.

Installing Prebuilt Distributions (Recommended)

Linux

The instructions below reference $XDG_DATA_HOME as an environment variable, it is recommended to export XDG variables in general, but if you have not, just replace the usages of $XDG_DATA_HOME below with $HOME/.local/share, which is the usual default for this XDG variable.

mkdir -p $XDG_DATA_HOME/llvm/
cd $XDG_DATA_HOME/llvm/
wget https://github.com/lumen/llvm-project/releases/download/firefly-15.0.0-dev_2022-07-22/clang+llvm-15.0.0-x86_64-linux-gnu.tar.gz
tar -xz --strip-components 1 -f clang+llvm-15.0.0-x86_64-linux-gnu.tar.gz
rm clang+llvm-15.0.0-x86_64-linux-gnu.tar.gz
cd -

MacOS

mkdir -p $XDG_DATA_HOME/llvm/
cd $XDG_DATA_HOME/llvm/
wget https://github.com/lumen/llvm-project/releases/download/firefly-15.0.0-dev_2022-07-22/clang+llvm-15.0.0-x86_64-apple-darwin21.5.0.tar.gz
tar -xzf clang+llvm-15.0.0-x86_64-apple-darwin21.5.0.tar.gz
rm clang+llvm-15.0.0-x86_64-apple-darwin21.5.0.tar.gz
mv clang+llvm-15.0.0-x86_64-apple-darwin21.5.0 firefly
cd -

Other

We don't yet provide prebuilt packages for other operating systems, you'll need to build from source following the directions below.

Building From Source

LLVM requires CMake, a C/C++ compiler, and Python. It is highly recommended that you also install Ninja and CCache to make the build significantly faster, especially on subsequent rebuilds. You can find all of these dependencies in your system package manager, including Homebrew on macOS.

We have the build more or less fully automated, just three simple steps:

git clone https://github.com/lumen/llvm-project
cd llvm-project
git checkout firefly
make llvm-shared

This will install LLVM to $XDG_DATA_HOME/llvm/firefly, or $HOME/.local/share/llvm/firefly, if $XDG_DATA_HOME is not set. It assumes that Ninja and CCache are installed, but you can customize the llvm target in the Makefile to use make instead by removing -G Ninja from the invocation of cmake, likewise you can change the setting to use CCache by removing that option as well.

NOTE: Building LLVM the first time will take a long time, so grab a coffee, smoke 'em if you got 'em, etc.

Building Firefly

Once LLVM is installed/built, you can build the firefly executable:

LLVM_PREFIX=$XDG_DATA_HOME/llvm/firefly FIREFLY_BUILD_TYPE=static cargo make firefly

This will create the compiler executable and associated toolchain for the host machine under bin in the root of the project. You can invoke firefly via the symlink bin/firefly, e.g.:

bin/firefly --help

You can compile an Erlang file to an executable (currently only on x86_64/AArch64):

bin/firefly compile [<path/to/file_or_directory>..]

This will produce an executable with the same name as the source file in the current working directory with no extension (except on Windows, where it will have the .exe extension).

NOTE: Firefly is still in a very experimental stage of development, so stability is not guaranteed.

Project Structure

Firefly is currently divided into three major components:

Compiler
Libraries
Runtimes

There are some crates in the root of the project that are in the process of being cleaned up/removed, so for the most part, the crates in compiler/, library/ and runtimes/ are those of interest.

Compiler

The Firefly compiler is composed of many small components, but the few most interesting are:

firefly_compiler, handles orchestrating the compiler itself, if you are looking for the driver of the compiler, this is it.
firefly_syntax_base, contains common types and metadata used across multiple stages of the compiler
firefly_syntax_erl, contains the abstract syntax tree, grammar, parser, and passes for performing semantic analysis, transformations on the AST, and lowering to Core. This is the primary frontend of the compiler, as it handles with user-provided Erlang code.
firefly_syntax_core, defines an intermediate representation that is based on an extended lambda calculus form, this is where a number of initial normalizations/transformations occur and corresponds to Core Erlang in the BEAM compiler.
firefly_syntax_kernel, defines an intermediate representation that is tailored towards pattern match compilation and code generation, it is flat relative to Core, funs/closures have been lifted, all variables have been made unique within their containing function, pattern matching has been compiled, and all calls have been transformed into static form and candiates for tail call optimization have been identified.
firefly_syntax_ssa, defines an SSA intermediate representation that is used for code generation; once transformed into SSA, performing codegen is very straightforward.
firefly_codegen, handles code generation from our SSA IR using MLIR/LLVM, and also contains the code responsible for linking objects/libraries/executables.

The other crates are all important as well, but are much smaller and tailored to a specific task, and so should be straightforward to grasp their function.

Libraries

There are a number of core libraries that are used by the runtime, but are also in some cases shared with the compiler (namely firefly_binary and firefly_number). These are designed to either be optional components, or part of a tiered system of crates that build up functionality for the various runtime crates.

firefly_system, provides abstractions over platform-specific implementation details that most of the runtime code doesn't need to know about. This primarily handles unifying low-level platform APIs.
firefly_alloc, provides abstractions for memory allocation and types which are allocator-aware, this is where our GC primitives are defined, as well as useful constructs like heap fragments.
firefly_rt, this is the primary core runtime library, hence the name, and provides the implementations of all the term types and their native APIs, as well as establishing things like the calling convention for Erlang functions, exceptions and backtraces, and other universal runtime concerns that cannot be delegated to a higher-level runtime crate.
firefly_arena, this is a helper crate that provides an implementation of a typed arena used in both the runtime and the compiler.
firefly_binary, this crate provides all the pieces for implementing binaries/bitstrings, including pattern matching primitives and constructors.
firefly_number, this crate provides the internal implementation of numeric types for both the compiler and runtime
firefly_beam, this crate provides native APIs for working with BEAM files, largely unused at the moment

Runtimes

The runtime is intended to be pluggable, but some parts are intended to always be included alongside those libraries:

firefly_crt, contains the primary entry point of Firefly-compiled executables, and is responsible for setting up the atom table, and the symbol table for dynamic dispatch.
firefly_tiny, contains our experimental runtime for development work

We have more robust runtime libraries that much time was invested into, but those are currently being reworked now that the compiler is done:

firefly_core, contains functionality useful across high-level runtimes, e.g. timer wheels
firefly_minimal, a richer version of firefly_tiny used with a previous iteration of the compiler

The above collection of libraries correspond to ERTS in the BEAM virtual machine.

Making Changes

At this stage of the project, it is important that any changes you wish to contribute are communicated with us first, as there is a good chance we are working on those areas of the code, or have plans around them that will impact your implementation. Please open an issue tagged appropriately based on the part of the project you are contributing to, with an explanation of the issue/change and how you'd like to approach implementation. If there are no conflicts/everything looks good, we'll make sure to avoid stepping on your toes and provide any help you need.

For smaller changes/bug fixes, feel free to open an issue first if you are new to the project and want some guidance on working through the fix. Otherwise, it is acceptable to just open a PR directly with your fix, and let the review happen there.

Always feel free to open issues for bugs, and even perceived issues or questions, as they can be a useful resource for others; but please do make sure to use the search function to avoid duplication!

If you plan to participate in discussions, or contribute to the project, be aware that this project will not tolerate abuse of any kind against other members of the community; if you feel that someone is being abusive or inappropriate, please contact one of the core team members directly (or all of us). We want to foster an environment where people both new and experienced feel welcomed, can have their questions answered, and hopefully work together to make this project better!

About Firefly

Firefly is not only a compiler, but a runtime as well. It consists of two parts:

A compiler for Erlang to native code for a given target (x86, ARM, WebAssembly)
An Erlang runtime, implemented in Rust, which provides the core functionality needed to implement OTP

The primary motivator for Firefly's development was the ability to compile Elixir applications that could target WebAssembly, enabling use of Elixir as a language for frontend development. It is also possible to use Firefly to target other platforms as well, by producing self-contained executables on platforms such as x86.

Firefly is different than BEAM in the following ways:

It is an ahead-of-time compiler, rather than a virtual machine that operates on bytecode
It has some additional restrictions to allow more powerful optimizations to take place, in particular hot code reloading is not supported
The runtime library provided by Firefly is written in Rust, and while very similar, differs in mostly transparent ways. One of the goals is to provide a better foundation for learning how the BEAM runtime is implemented, and to take advantage of Rust's more powerful static analysis to catch bugs early.
It is designed to support targeting WebAssembly, as well as many other types of targets.

The result of compiling a BEAM application via Firefly is a static executable. This differs significantly from how deployment on the BEAM works today (i.e. via OTP releases). While we sacrifice the ability to perform hot upgrades/downgrades, we make huge gains in cross-platform compatibility, and ease of use. Simply drop the executable on a compatible platform, and run it, no tools required, or special considerations during builds. This works the same way that building Rust or Go applications works today.

Goals

Support WebAssembly/embedded systems as a first-class platforms
Produce easy-to-deploy static executables as build artifacts
Integrate with tooling provided by BEAM languages
More efficient execution by removing the need for an interpreter at runtime
Feature parity with mainline OTP (with exception of the non-goals listed below)

Non-Goals

Support for hot upgrades/downgrades
Support for dynamic code loading

Firefly is an alternative implementation of Erlang/OTP, so as a result it is not as battle tested, or necessarily as performant as the BEAM itself. Until we have a chance to run some benchmarks, it is hard to know what the difference between the two in terms of performance actually is.

Firefly is not intended to replace BEAM at this point in time. At a minimum, the stated non-goals of this project mean that for at least some percentage of projects, some required functionality would be missing. However, it is meant to be a drop-in replacement for applications which are better served by its feature set.

Architecture

Compiler

The compiler frontend accepts Erlang source files. This is parsed into an abstract syntax tree, then lowered through four middle tiers where different types of analysis, transformation, or optimization are performed:

Core IR (similar to Core Erlang)
Kernel IR (similar to Kernel Erlang)
SSA IR (a transformation of Kernel IR in preparation for codegen)
MLIR (the final stage where optimizations and certain transformations occur)

The final stage of the compiler lowers MLIR to LLVM IR and then LLVM handles generating object files from that. Our linker then takes those object files and produces a shared library or executable (the default).

In MLIR, and particularly during the lowering to LLVM IR, all high-level abstractions around certain operations are stripped away and platform-specific details take shape. For example, on x86_64/AArch64, hand-written assembly is used to perform extremely cheap stack switching by the scheduler, and to provide dynamic function application facilities for the implementation of apply.

Runtime

The runtime design is mostly the same as OTP, but we are not running an interpreter, instead the code is ahead-of-time compiled:

The entry point sets up the environment, and starts the scheduler
The scheduler is composed of one scheduler per thread
Each scheduler can steal work from other schedulers if it is short on work
Processes are spawned on the same scheduler as the process they are spawned from, but a scheduler is able to steal them away to load balance
I/O is asynchronous, with dedicated threads and an event loop for dispatch

The initial version will be quite spartan, but this is so we can focus on getting the runtime behavior rock solid before we circle back to add in more capabilities.

NIFs

Currently it is straightforward to define NIFs in Rust without the overhead of erl_nif, but we don't yet have an abstraction that allows us to take existing NIFs designed around erl_nif and make them work. This is something in the pipeline, but is not yet a high priority for us.

License

Apache 2.0