You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When rayon's par_iter() is used, none of the optick::event!() calls are logged from the spawned threads. I'm curious if this is an issue on the optick side or the rayon side, would love someone more knowledgeable to take a look.
If optick is using thread-specific state, then it wouldn't surprise me at all that it fails to see inside the rayon thread-pool. I've never used it, but at a glance optick::start_capture() does look like it's only for the current thread.
You might be able to use a ThreadPoolBuilder::start_handler to turn it on, but it will probably still get confused about the hierarchy of events due to rayon's work-stealing. See also #915.
I made a sample repo (that contains a sample capture) to demonstrate the issue - https://github.com/vladkvit/test_rayon_optick
When rayon's par_iter() is used, none of the optick::event!() calls are logged from the spawned threads. I'm curious if this is an issue on the optick side or the rayon side, would love someone more knowledgeable to take a look.
FWIW, I also cross-posted the issue here - bombomby/optick-rs#9
The text was updated successfully, but these errors were encountered: