Abhisheklearn12 commented on PR #25793:
URL: https://github.com/apache/datafusion/pull/25793#issuecomment-5864780085

   > @Abhisheklearn12 Could you check the affected ARM binaries to confirm 
whether this is code placement, and whether it comes from the production change 
or the added benchmark cases?
   
   hi, @kumarUjjawal i checked on aarch64 by reproducing the bot's build (same 
`cargo bench` command, thin lto, rust 1.98.1) for main, this pr, prod change 
only, and bench cases only. repeated with different crate hashes as a control, 
since a cross build can't match the bot's native ones.
   
   - the functions behind the slower rows have identical instructions in every 
variant. only branch targets and data addresses differ.
   - both changes shift them: prod by ~6-10kb, the new bench cases by ~400b. 
either alone lands them at a different cache line offset.
   
   so it's code placement from both changes. the exact offsets in the bot's 
binary depend on crate hashes, and i couldn't 
   time on arm.


-- 
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