📦 Bundle size

Full ESM build, minified and gzipped.

🌳 Bundle size after tree-shaking

Minimal build including a renderer, camera, empty scene, and dependencies.

I believe that with this PR we can't do:

The usual WGSL syntax:

const a = atomicAdd(buffer.element(0), 1).toConst('a')

nor:

const a = uint(0).toVar('a')
atomicAdd(buffer.element(0), 1, a)

/cc @sunag

Hmm.. I think the usage of atomicAdd should keep the same with this PR or not?

Not really, in most case with atomic operations we need to access the value of the operation while doing it (see my syntax in my previous message), for example doing an operation in a loop while trying to access to that value updated.

It's currently not possible anymore, only vagabond operation which is very limiting.

I still don't understand your example exactly, but I believe you are referring to the function call not being added to the Stack immediately?

Ah sorry about that, I should have provided a better example. But basically from the specs:
https://www.w3.org/TR/WGSL/#atomic-rmw

Each function returns the original value stored in the atomic object before the operation.

For example:

fn atomicAdd(atomic_ptr: ptr<AS, atomic<T>, read_write>, v: T) -> T

Which means that any atomic operation will return a value, so it would be great to be able to easily read the returned value, for example:

The usual WGSL syntax:

const a = atomicAdd(buffer.element(0), 1).toConst('a')

nor:

const a = uint(0).toVar('a')
atomicAdd(buffer.element(0), 1, a)

This was possible to do before this PR via:

const a = uint(0).toVar('a')
atomicAdd(buffer.element(0), 1, a)

but not anymore.

Seconding @RenaudRohlinger, this update broke some of my projects because I can't access the return value of the atomic operation anymore.

I'm making a fix for this, just need to do a few more checks.