Does anyone know the target triple of wasmJs? is i...

# webassembly

[Charlie Tapping]

Does anyone know the target triple of wasmJs? is it wasm32-unknown-unknown?

[Svyatoslav Kuzmich [JB]]

Are you referring to LLVM triple? wasmJs doesn’t have it because it doesn’t use LLVM.

[Charlie Tapping]

It doesn’t? surely its lowered using llvm? or its a custom compiler backend?

[Svyatoslav Kuzmich [JB]]

It is a custom compiler backend that generates wasm directly. Binaryen is then used as a wasm->wasm post-processing optimizer pass. We used to have experimental LLVM-based Kotlin/Native wasm32 target, but it is removed now.

[Charlie Tapping]

interesting, do you happen to know why they opted for that route? I’d hazard a guess that it would allow them to experiment with certain proposals but llvm moves very fast itself?

[Svyatoslav Kuzmich [JB]]

(I’m one of those who started making the new backend)
There were a few reasons:
1. Wasm GC proposal allows us to interact with JS and DOM GC. It is a built-in advanced GC, that allows us to make binary size smaller.
2. Wasm with a GC proposal is a higher level of abstraction compared to LLVM IR.
It was unclear at the time if LLVM project would be OK with added complexity and support it (it still hasn’t). Even with LLVM support, we would likely need to generate a different version of LLVM IR for wasm.
3. We can acheive faster build times when targeting a single platform.

[Svyatoslav Kuzmich [JB]]

See https://youtrack.jetbrains.com/issue/KT-46773/Implement-an-experimental-version-of-the-Kotlin-Wasm-compiler-backend for more context

[Charlie Tapping]

That makes a lot of sense thank you!

[Charlie Tapping]

I was originally asking this because I had guessed we we’re using lllvm and the web target was empscripten. Is there any plans to bring c interop support to wasm? It would mean I could share more of my KMP project without having to write bespoke impls

[Svyatoslav Kuzmich [JB]]

No short-term plans due to priority in other areas.

We see quite a bit of interest in reusing C libraries in Kotlin. Current solutions like JNI and Kotlin/Native C interop allow this, but are platform-specific. I, personally, would love to see some multi-platform solution (KT-39144) in the future.

In the meantime, we want to have some form of C interop in Wasm. Some use cases will be covered when we support Component Model. It is unclear to me how good this approach will be in practice. If it works well in Wasm, I could imagine it becoming a multi-platform solution for foreign code sharing. If it doesn’t work, we’ll likely want some dedicated C interop solution (at least Wasm platform specific).