perf: improve ipc poll logic #4037
Conversation
Codecov Report
... and 9 files with indirect coverage changes
Flags with carried forward coverage won't be shown. Click here to find out more.
|
rkrasiuk
added
the
C-perf
| let mut budget = 5; | ||
|
|
||
| // ensure we still have enough budget for another iteration | ||
| 'outer: loop { | ||
| budget -= 1; | ||
| if budget == 0 { | ||
| // make sure we're woken up again | ||
| cx.waker().wake_by_ref(); | ||
| return Poll::Pending | ||
| } |
There was a problem hiding this comment.
not quite clear to me - why do we need to return early? is it because the task is blocking and you want to yield back control to the scheduler?
There was a problem hiding this comment.
yes, this ensures this loops doesn't run for too long, since this is polled concurrently with something that also receives responses (from subscriptions)
| 'inner: loop { | ||
| let mut drained = false; | ||
| // drain all calls that are ready and put them in the output item queue | ||
| if !this.pending_calls.is_empty() { |
There was a problem hiding this comment.
these are already effectively bounded by the rate limit params that restrict how many calls can be active
| Poll::Ready(res) => { | ||
| let item = match res { | ||
| Ok(Some(resp)) => resp, | ||
| Ok(None) => continue 'inner, |
There was a problem hiding this comment.
under which circumstances this can return Ok(None)?
There was a problem hiding this comment.
for subscriptions responses which are routed differently
improve poll logic of ipc connection
this makes it slightly more complicated but now priorities output over input, since output is the real bottleneck if input is high.
ref #4028