pnowojski opened a new pull request #12267:
URL: https://github.com/apache/flink/pull/12267
For some reason the following commit:
54155744bd [FLINK-17547][task] Use RefCountedFile in SpanningWrapper
caused a performance regression in various benchmarks. It's hard to tell why
as none of the benchmarks are using spill files (records are too small), so
our best guess is that combination of AtomicInteger inside RefCountedFile
plus NullPointerException handling messed up with JIT ability to get rid
of the memory barrier (from AtomicInteger) on the hot path.
## Verifying this change
This change is covered by existing micro benchmarks.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): (yes / **no**)
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: (yes / **no**)
- The serializers: (yes / **no** / don't know)
- The runtime per-record code paths (performance sensitive): (**yes** / no
/ don't know)
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Kubernetes/Yarn/Mesos, ZooKeeper: (yes / **no** /
don't know)
- The S3 file system connector: (yes / **no** / don't know)
## Documentation
- Does this pull request introduce a new feature? (yes / **no**)
- If yes, how is the feature documented? (**not applicable** / docs /
JavaDocs / not documented)
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
