leostera / lam

:rocket: a lightweight, universal actor-model vm for writing scalable and reliable applications that run natively and on WebAssembly

 

LAM is a lightweight, universal virtual machine for writing scalable and reliable applications that run natively and on WebAssembly.

It is inspired by Erlang and Lua, and it is compatible with the Erlang VM.

LAM lets you reuse the same programming paradigm, known for being productive, across your entire application stack.

Come join us on Discord! (Ps: we share a server with Caramel)

Features

• Runs Natively and on WebAssembly -- pick and choose your runtime!
• Easy to Target -- a small and specified bytecode with a text and binary format
• Erlang VM compatibility -- run your existing Erlang, Elixir, Caramel, and Gleam code
• Seamless multi-core -- built to scale from one to thousands of cores for free
• Extreme reliability -- use Erlang's OTP supervision patterns

Status

Still under heavy development!

There's plenty of work to be done for it to be fully usable, but we keep a few tracking issues here:

• LAM Specification Status
• Targetability Status
• WebAssembly and Native Runtime Support

The Erlang and Elixir ecosystem compatibility is tracked here:

• Erlang VM Compatibility Status
• Erlang/Elixir Support Across Runtimes

Getting Started

You can download the latest binary from the releases page. After unpacking it you should be able to add it to your PATH env and start playing around with the lam binary.

Like this:

# in this example I'm running linux with glibc
$ wget https://github.com/AbstractMachinesLab/lam/releases/download/v0.0.7/lam-v0.0.7-x86_64-unknown-linux-gnu.tar.gz
$ tar xzf lam-*
$ export PATH=$(pwd)/lam/bin:$PATH

Now we can do a quick test. Make a file test.erl with this contents:

-module(test).

-export([main/1]).

main([]) -> ok;
main([Name|T]) ->
  io:format(<<"Hello, ~p!\n">>, [Name]),
  main(T).

And we can compile it to BEAM byte code and use LAM to build a binary for it, like this:

$ erlc test.erl
$ lam build test.beam --output test.exe --target native --entrypoint test
$ ./test.exe Joe Robert Mike
Hello, Joe!
Hello, Robert!
Hello, Mike!

How does it work?

LAM compiles your .beam files ahead of time into a representation that's optimized for running them.

Then it bundles that with the appropriate target runtime into some binary output.

                       binary
                    instructions
                   +------------+              output
 .beam files +---->| 1001101110 |-----+    +-----------+
                   +------------+     |    | .exe      |
                                      |--->| .wasm     |
                   +-------------+    |    | .wasm/.js |
                   | LAM RunTime |----+    +-----------+
                   +-------------+