Rich-T-kid commented on issue #1274:
URL: https://github.com/apache/arrow-rs/issues/1274#issuecomment-5564714415

   This seems pretty straightforward to implement, I can start working on this.
   
   I guess this most important piece of this issue is determining what 
benchmarks we'd like to use as a guage betweeen versions. In datafusion this is 
an easier question to answer since there are standards already out 
(clickbench,TPCH/10...ect).
   
   Id reccomend 
   1. filter kernels (`take`,`filter_record_batch`)
   2. BatchCoalescer kernels (RepartionExec)
   3. concatenate kernels 
   4. row format kernels  (Aggregations/sorts)
   5. sort kernel 
   6. cast kernel
   7. ipc kernel  (flushing to disk)
   
   due to their direct ties to query exection in datafusion downstream
   
   
   then we'd also have the parquet related kernels 
   
   1. parquet reader kernel
   2. parquet writer kernel
   
   Im personally not super familiar with all the sub-sections of the parquet 
kernel so any other reqs would be useful.
   
   It may also be nice to include the side by side differneces between major 
releases as well, similar to what @alamb does for [datafusion major releases 
](https://datafusion.apache.org/blog/output/2026/08/25/datafusion-55.0.0/#appendix-clickbench-results).
   
   
   I'll let this sit for a few days in case anyone has opions on which kenerls 
to include/leave out then ill open a PR 🚀 
   
   


-- 
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: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to