lichuang commented on PR #12356:
URL: https://github.com/apache/arrow-rs/pull/12356#issuecomment-7055085762

   Thanks @alamb — agreed that the gain here is visible by inspection, and I'm 
not trying to argue the change needs a benchmark to be justified.
   
   The concrete problem is the automated runner. `run benchmark` needs the 
benchmark to exist on the base revision. #12228 was reviewed without a 
`builder_finish` benchmark on `main`, so the bot run could not measure it at 
all and the evidence stayed local. This PR exists so the follow-up builder 
changes can be compared base-vs-head by the runner instead of by hand.
   
   If the general worry is a proliferation of micro-benchmarks chasing 
allocator noise: that's fair, and I'm happy to limit this to the per-`finish` 
shape, which is where the `ArrayData` overhead actually shows up, and to drop 
any group you think is not pulling its weight.
   
   I'm fine either way — if you'd rather skip the benchmark and land the 
builder changes with inspection-only justification, I'll rebase #12353 and drop 
this 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