kanaka/warpy

WebAssembly interpreter in RPython

warpy - WebAssembly in RPython

A WebAssembly interpreter written in RPython.

Warpy supports the WebAssembly MVP (minimum viable product) version of the spec.

Prerequisites

Build whe warpy executable. You will need a build of rpython on the path:

make warpy-jit   # or make warpy-nojit

Alternatively, you can use an rypthon docker container (built from Dockerfile-rpython) and do the compilation from there:

docker pull kanaka/warpy-rpython
docker run -it kanaka/warpy-rpython -v `pwd`:/build -w /build make warpy-jit

You will need wast2wasm to compile wast source to wasm bytecode. Check-out and build wabt (wabbit):

git clone --recursive https://github.com/WebAssembly/wabt
make -C wabt gcc-release

Usage

Compile a wasm module:

wast2wasm test/addTwo.wast -o test/addTwo.wasm

Load and call a function in a wasm module:

./warpy-jit test/addTwo.wasm addTwo 11 12

You can also use standard python (but it's much slower of course):

python warpy.py test/addTwo.wasm addTwo 13 14

There is also a REPL mode that allow you to interactively invoke functions within a module:

./warpy-jit --repl test/addTwo.wasm
webassembly> addTwo 2 3

Misc

Some rough notes for running the WebAssembly specification tests can be found in notes\_testing.

License

MPL-2.0 (see LICENSE)

warpy's Issues

Idea: use the wasm test suite

wasm has a test suite which can be used to trigger a lot of corner cases. As an example, we used in in ppci like this:

https://github.com/windelbouwman/ppci-mirror/blob/master/test/suite_adapters/test_wasm_spec_suite.py

This is not really an issue, but more an idea?

Change `wast` to `wat` in README.md

From README.md

Compile a wasm module:
wast2wasm test/addTwo.wast -o test/addTwo.wasm

I found no wast file in test folder. They are wat files.

Also, the tool name is wat2wasm, not wast2wasm

Thanks!

is there a way to import a python object in the wasm memory/table?

Extending warpy to allow for full state dump

I'm interested in building a WebAssembly interpreter that supports not only dumping the memory, globals and tables, but the full stack state as well. Do you think it is sensible to build on your project to achieve this?

The overhead may wipe out some JIT advantages, but as long as its not worse than a 10x slowdown, I'm fine with that.

Edit: Oh, I see warpy already has tracing built in! Sweet!