Built-in atomic functions added

[nidin]

Built-in atomic functions added for SharedArrayBuffer.

List of functions

Atomic.load<T>(offset: usize, constantOffset?: usize): T
Atomic.store<T>(offset: usize, value: void, constantOffset?: usize): void
Atomic.add<T>(ptr: usize, value: T, constantOffset?: usize): T
Atomic.sub<T>(ptr: usize, value: T, constantOffset?: usize): T
Atomic.and<T>(ptr: usize, value: T, constantOffset?: usize): T
Atomic.or<T>(ptr: usize, value: T, constantOffset?: usize): T
Atomic.xor<T>(ptr: usize, value: T, constantOffset?: usize): T
Atomic.xchg<T>(ptr: usize, value: T, constantOffset?: usize): T
Atomic.cmpxchg<T>(ptr: usize, expected:T, replacement: T, constantOffset?: usize): T
Atomic.wait<T>(ptr: usize, expected:T, timeout:i64): i32
Atomic.notify<T>(ptr: usize, count: u32): u32

[dcodeIO]

elementsLookup should also contain namespace members here, like both Atomic and Atomic.load etc. Is there a bug?

[nidin]

I will cross check. Atomic.load was not in elementsLookup but it was a few months ago.

[nidin]

For some reason namespace members are undefined in elementsLookup

compiler.program.elementsLookup.get("Atomic.load")

return undefined

---

compiler.program.elementsLookup.get("Atomic")

return

MapIterator {
  'load',
  'store',
  'add',
  'sub',
  'and',
  'or',
  'xor',
  'xchg',
  'cmpxchg',
  'wait',
  'notify' }

[MaxGraey]

It seems typeArguments[0] used 5 times so it will be great cache it to variable and may be in upper scope

[nidin]

I will change it.

[MaxGraey]

.toString(10) is unnecessary. Just use toString()

[nidin]

This too.

[MaxGraey]

btw you could try turn on threads for latest node via node --experimental-wasm-threads and write some tests.

[willemneal]

How do threads work with node? In the Browser there are web workers, but I thought node was single threaded.

[MaxGraey]

Mostly the same for now: https://nodejs.org/api/worker_threads.html

[nidin]

@MaxGraey This PR only includes built-in functions to make review easy. PR #278 Contains everything but it's too much for a single review. I am actually breaking the PR #278 into small parts.

[MaxGraey]

I see. Totally makes sense. But also will great have small test cases of functionality present in this PR. Or add it in separate PR

[nidin]

Atomic operations won't compile without shared memory support in compiler. It will come in next PR. followed by test cases.

[MaxGraey]

Got it!

[dcodeIO]

Great, thanks! :)