Dandandan commented on PR #21484:
URL: https://github.com/apache/datafusion/pull/21484#issuecomment-4215771252

   > I suspect the branch predictor will figure this out pretty fast since it 
doesn't change, and the dominant cost is the virtual call through Box<dyn Fn>. 
I can try that out, but don't anticipate much of a change. I'd be hypocrite 
based on what I said above though to not specialize for single-columns if I 
believe multi-column is rare.
   
   The branch predictor will work well but the generated code will be scalar / 
not vectorized and needs to execute much more instructions per item.
   I expect at least when writing a benchmark for the operation itself it will 
help, not so sure if it is still a bottleneck for SMJ?


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to