mapleFU commented on issue #39365:
URL: https://github.com/apache/arrow/issues/39365#issuecomment-1871691424

   Yeah, batchSize would be help, usally we hope batches can be staying in L1 
cache. And larger batch size can have smaller cost.
   
   If the computation is complex and has many operators (like group by. ... 
etc), writing a self generated call is complex, but workload like 
`(u1.d1.s1*u1.d2.s4+u1.d2.s4/u1.d1.s1)/u1.d1.s1*u1.d2.s4` might benifits from 
less memory allocation and fast computation, so it depend on your workload


-- 
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