Stream: wasmtime

Topic: Pulley specification


view this post on Zulip Julien Cretin (ia0) (Oct 21 2024 at 20:44):

Is there a timeline on when the Pulley bytecode would be somehow stable? I'm asking in case I can't compile Pulley interpreter to my target and need to write my own interpreter.

view this post on Zulip Chris Fallin (Oct 21 2024 at 20:45):

I believe the intent is for there to be one Pulley interpreter, the one that we provide; the code that it runs will depend on the rest of Wasmtime's runtime in any case. Could you clarify the use-case you have in mind?

view this post on Zulip Julien Cretin (ia0) (Oct 21 2024 at 20:53):

The context is Wasefire, essentially a way to run Wasm applets on embedded devices. Currently it has a very basic in-place Wasm interpreter (very low memory overhead). But it's slow. Pulley would improve the situation.

view this post on Zulip Chris Fallin (Oct 21 2024 at 20:55):

Cool, makes sense. So yeah the way to think of this is that Wasmtime + its Pulley backend is together a replacement for a Wasm interpreter (with optional AOT preparation step re: Pulley bytecode)

view this post on Zulip Julien Cretin (ia0) (Oct 21 2024 at 20:59):

Yes, modulo the validation step that is currently done on device and will have to move out (requiring some mechanism to prove that the compiled Pulley bytecode comes from a valid Wasm module, but I'll anyway need such a mechanism for other reasons)


Last updated: Dec 23 2024 at 13:07 UTC