Improve concurrent task runner performance #6948
Merged
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Outdated suggestions cannot be applied.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.
Improves performance of the concurrent task runner by using events instead of spinlocks to wait for results where feasible. I believe this only doesn't apply to use of async tasks from sync flows. If performance is a major concern in that case, users should be using sync or async consistently. This follows a pattern established in
PrefectFuture.wait
.Closes #6468
Supersedes #6239
Example
In #6468 the concurrent task runner was demonstrated to be 7233% slower when performing synchronous IO. Replicated here, the concurrent task runner is 16% faster when performing synchronous IO. I did not attempt to reproduce the original slow download as my internet is not as fast as theirs and their example would take a theoretical 3 hours to download on my machine.
I failed to demonstrate an improvement in total flow runtime with a benchmark using CPU intensive tasks derived from #6239.
Checklist
<link to issue>
"fix
,feature
,enhancement