This page is a work-in-progress and will be updated as the project develops.
For now, the calendar will just reflect dates for the 2017 Summer Term in accordance with the Haskell Summer of Code schedule (which this project will hopefully be a part of)
We got accepted to HSoC! Check out the blog for more
This project will consist of three main phases.
- Extending GHC to produce WASM-compatible LLVM.
- Ensuring the WebAssembly generated by LLVM is valid and callable from the browser.
- Porting the GHC runtime system.
Phase one will involve defining a WebAssembly target in GHC for emitting LLVM, as well as finalizing the necessary nix expressions to build this extended version of GHC.
Phase two features two major potential issues. One, LLVM’s support for tailcalls and the GHC calling convention. In addition, it’s currently unclear whether LLVM properly handles these things when targeting WebAssembly; WebAssembly doesn’t implement tailcalls, and the GHC calling convention is very unusual.
Phase three is likely to be the most involved stage of the project. Parts of the RTS written in C may, with the aid of LLVM compilation tools, work without much additional work. Other parts of the RTS written in Cmm should be able to be compiled with the help of the GHC LLVM backend. Various components of the RTS (syscalls, etc.) are likely to need special attention.
The minimum viable product for this project is a tool that can produce valid runnable WebAssembly from Haskell code. It will support an unthreaded runtime, awaiting WebAssembly’s threading implementation for the threaded runtime. The C FFI will be implemented via the same means as the C FFI with a native backend. Only essential syscalls will be implemented.
We plan to have a version of the GHC toolchain that has knowledge of an available, but non-functioning, WebAssembly backend by July 2. By the week of July 23, the project should be able to produce valid WebAssembly from Haskell code. The rest of the term will be spent porting the GHC RTS. From our preliminary research, we believe the first two main objectives should be able to be completed reasonably quickly (possibly ahead of schedule). The bulk of this project will then be spent on the final main objective.
In the following calendar, goals are intended to be completed by the week they are listed in. Main Objectives are in bold
|June 18||Project website ported to Hakyll with improved layout|
|July 2||GHC building with WASM as a Target|
|July 23||WebGHC producing valid WASM|
|August 27||RTS working in WASM|