Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Soften language for notify on unshared memories #191

Merged
merged 1 commit into from
Sep 13, 2022
Merged

Conversation

tlively
Copy link
Member

@tlively tlively commented Sep 12, 2022

Since #144, we've specified that memory.atomic.notify is allowed to be executed on unshared memories, but always returns 0 since it is impossible for there to be anything waiting on an unshared memory since all variations of the wait instructions in Wasm and JS trap or throw exceptions on unshared memories. However, I realized recently that there is no technical reason why Atomics.waitAsync shouldn't be able to wait on an unshared memory. Because Atomics.waitAsync returns a promise rather than blocking, it can in principle be used to synchronize between multiple asynchronous tasks in flight concurrently in the same JS context, and now that JSPI is becoming more real, this actually might be a useful thing to do. In particular Atomics.waitAsync could be used to build suspending versions of mutexes and condition variables to synchronize between threads running as different asynchronous tasks within a single JS context.

Actually realizing this vision of asynchronous threading with JSPI on top of Atomics.waitAsync requires changes on the JS side to allow Atomics.waitAsync and Atomics.notify to be used on unshared ArrayBuffers, but in the meantime we can prepare for that change on the Wasm side by softening the language stating that memory.atomic.notify unconditionally returns 0 for unshared memories.

cc @syg

Since #144, we've specified that `memory.atomic.notify` is allowed to be executed on unshared memories, but always returns 0 since it is impossible for there to be anything waiting on an unshared memory since all variations of the wait instructions in Wasm and JS trap or throw exceptions on unshared memories. However, I realized recently that there is no technical reason why [`Atomics.waitAsync`](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Atomics/waitAsync) shouldn't be able to wait on an unshared memory. Because `Atomics.waitAsync` returns a promise rather than blocking, it can in principle be used to synchronize between multiple asynchronous tasks in flight concurrently in the same JS context, and now that [JSPI](https://github.com/WebAssembly/js-promise-integration/blob/main/proposals/js-promise-integration/Overview.md) is becoming more real, this actually might be a useful thing to do. In particular `Atomics.waitAsync` could be used to build suspending versions of mutexes and condition variables to synchronize between threads running as different asynchronous tasks within a single JS context.

Actually realizing this vision of asynchronous threading with JSPI on top of `Atomics.waitAsync` requires changes on the JS side to allow `Atomics.waitAsync` and `Atomics.notify` to be used on unshared ArrayBuffers, but in the meantime we can prepare for that change on the Wasm side by softening the language stating that `memory.atomic.notify` unconditionally returns 0 for unshared memories.
Copy link
Collaborator

@conrad-watt conrad-watt left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Makes sense, thanks!

@conrad-watt conrad-watt merged commit 7fb4187 into main Sep 13, 2022
@conrad-watt
Copy link
Collaborator

(I'd also be interested in @syg's take on how likely this is on the JS side)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants