Bound channel elements processed every time process_events is called #200
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #200 +/- ##
==========================================
+ Coverage 85.08% 85.44% +0.36%
==========================================
Files 13 13
Lines 1837 1876 +39
==========================================
+ Hits 1563 1603 +40
+ Misses 274 273 -1 ☔ View full report in Codecov by Sentry. |
|
It appears this has exposed a bug in Windows' |
At the moment, the channel implentation processes as many channel elements as possible every time "process_events" is called. However, in multithreaded cases this can cause the loop to be stuck in the "process_events" section of a channel forever. If one thread keeps sending new elements into the channel while the current thread keeps reading them, it will starve other event sources of running time. This commit fixes this issue by bounding the number of channel elements that can be processed every time "process_events" is called. It chooses the smallest of the following numbers: - The capacity of the channel. - 1024 (chosen because this is also used by async-executor) If the channel is not empty after we have read this number of elements, the underlying source is not triggered. This should make it so the channel is immediately re-polled on the next dispatch. However it gives other sources more time to run. Signed-off-by: John Nunley <[email protected]>
Signed-off-by: John Nunley <[email protected]>
Previously, we only processed the source in the channel/executor implementations when we needed to clear the readiness; however this causes issues in the Windows `Ping` implementation. As a workaround, this commit always calls "process_events" in `Ping`, but simply re-notifies the PingSource if we need to try again. Signed-off-by: John Nunley <[email protected]>
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
At the moment, the channel implentation processes as many channel
elements as possible every time "process_events" is called. However,
in multithreaded cases this can cause the loop to be stuck in the
"process_events" section of a channel forever. If one thread keeps
sending new elements into the channel while the current thread keeps
reading them, it will starve other event sources of running time.
This commit fixes this issue by bounding the number of channel elements
that can be processed every time "process_events" is called. It chooses
the smallest of the following numbers:
If the channel is not empty after we have read this number of elements,
the underlying source is not triggered. This should make it so the
channel is immediately re-polled on the next dispatch. However it gives
other sources more time to run.