Add spinlock implementation without std::thread::sleep
#138
Conversation
zesterer
left a comment
zesterer
left a comment
There was a problem hiding this comment.
Thanks for the reminder, this PR flew under the radar.
|
Great, thanks for the PR! |
|
?? |
|
maybe a ci wasm build check would be nice? to guarantee that doesn't break again |
|
@zesterer ?? |
|
Sorry about that. I replied, ran CI, then didn't get back to this. Anyhow, much appreciated! |
|
Tysm!! Would it be possible to get this published on crates.io as well? |
|
@zesterer I'm sorry for constantly pinging you- but it seems to be the only way to get your attention? |
|
I should have some free time tomorrow to do this, thanks for the reminder. |
|
Not forgotten about this, trying to find the time to write up what turns out to be quite an extensive changelog! |
|
take your time, we appreciate your efforts :-) |
|
I've just released |
|
incredible. thanks! |
4 participants
This pull request adds aspin-plainfeature that uses a spinlock-based locking system like thespinfeature does, but without usingstd::thread::sleep. Closes #137.Edit: This pull request changes the
spinfeature to only usestd::thread::sleepon Unix and Windows. Closes #137.Some targets like
wasm32-unknown-unknowndon't support blocking, so you won't be able to use flume on those platforms without this new feature. Even if you enable-Ctarget-feature=+atomics,+bulk-memory,+mutable-globalson thewasm32-unknown-unknowntarget, flume still doesn't work on the main thread of a web browser, it only works in web workers. See #137 for more details about this, but essentially the problem is that thewait_lockfunction in flume has two different implementations and both of them are blocking, including the current spinlock-based one.