zanmato1984 opened a new issue, #45611:
URL: https://github.com/apache/arrow/issues/45611

   When benchmarking hash join build phase using [this 
benchmark](https://github.com/apache/arrow/blob/31994b5c2069a768e70fba16d1f521e4de64139e/cpp/src/arrow/acero/hash_join_benchmark.cc#L301),
 I saw significant contention when the concurrency is high. As the following 
flame graph shows, function 
[PartitionLocks::AcquirePartitionLock](https://github.com/apache/arrow/blob/31994b5c2069a768e70fba16d1f521e4de64139e/cpp/src/arrow/acero/partition_util.cc#L56)
 (the long yellow bar in the middle) takes quite a lot CPU time (thanks to that 
it is using spin lock rather than mutex, otherwise it may not be so 
eye-catching).
   
   We should consider optimizing this.
   
   
![Image](https://github.com/user-attachments/assets/16282918-b854-4807-86cc-239e1c53046f)


-- 
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.

To unsubscribe, e-mail: issues-unsubscr...@arrow.apache.org.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org

Reply via email to