shuttie commented on issue #10529: [FLINK-15171] [serialization] fix 
performance regression caused by too many buffer allocations on string 
serialization
URL: https://github.com/apache/flink/pull/10529#issuecomment-566594243
 
 
   Looks like I was over-optimistic: while running 
`SerializationFrameworkMiniBenchmarks` with this PR, regression is still here 
(but not as dramatic as before). While running the benchmark code with 
async-profiler and `-XX:+DebugNonSafepoints` enabled, I finally was able to 
find where is the difference between my benchmark and the 
`SerializationFrameworkMiniBenchmarks`: they have different underlying buffer 
implementations with really different performance characteristics.
   * `roundtripTuple` uses a thin wrapper on heap-only `byte[]`.
   * `SerializationFrameworkMiniBenchmarks` uses HybridMemorySegment, which is 
off-heap by default.
   
   I will try to reproduce the issue yet another time, but with different 
underlying buffer implementations.

----------------------------------------------------------------
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]


With regards,
Apache Git Services

Reply via email to