Please implement Atomics.waitAsync as well!

[tlively]

The stage 3 Atomics.waitAsync proposal combined with JSPI will allow Emscripten to stop busy-waiting for locks, condition variables, and other blocking synchronization primitives on the main thread, which will be a great win for performance and efficiency for large, multithreaded applications. It would be great if we could assume that the presence of JSPI implied the presence of Atomics.waitAsync across browsers so we don't have to maintain a polyfill or continue busy waiting.

[fgmccabe]

I don't think that I would support adding an explicit dependency here. What is the implementation status of atomics?

[tlively]

This is just an informal request for implementers to consider :)

Atomics.waitAsync is its own TC39 proposal, and while all browsers support general Atomics, only blink currently supports Atomics.waitAsync.

[ajklein]

Some updates:

• Atomics.waitAsync made it to Stage 4 and is now merged into the core JS spec
• It's supported in Safari 16.4
• Firefox seems to be the remaining non-supporting browser (tracking bug).

@eqrion FYI as you're thinking about prioritizing JSPI work.

[fgmccabe]

Current semantics of JSPI should address this question.