JF Bastien

That (flag + apply operation) seems feasible.

@lars-t-hansen does your implementation ever need to move the buffer to grow?

I think the final design should allow for implementations that require moving the memory, as well as for those that never move. I'm not 100% on what that would change,...

> After that it's likely that the signal handler will have to deal with the case where we observe a protection fault while some other thread is in the process...

I would rather see OpenMP support initially implemented as a custom thread pool using WebAssembly's primitives, i.e. zero VM work should need to be done to support OpenMP, only toolchain...

I'm not sure I understand what you're looking for exactly. IIUC you'd like to share a memory, but know that parts of it aren't modified? That could be done through...

> Its true that high-level code in a safe language like Rust is lowered to unsafe machine instructions for execution. The question I'm interested in is whether, as the module...

> I don't think @DLehenbauer suggests diverging from the current approach, rather it's about gaging interest in "a subset of wasm that is verifiably safe with respect to concurrency". I'm...

I'm not sure I understand why you think changing to i64 is a bad idea. Could you elaborate?

Yeah that sounds fine. I don't see anything bad with i64, except maybe a slight incompatibility with Number on the JS side (and JS could return inf on overflow...)