jhorstmann commented on PR #4441:
URL: https://github.com/apache/arrow-rs/pull/4441#issuecomment-1610027588

   > I know also of a few (cloud) users are enabling a specific cpu target
   
   We certainly do, especially in a cloud or server side environment we want to 
make the best use of the available hardware. So our usecase does not benefit 
from runtime dispatching, but I understand this might not be the main usecase. 
I'm also a huge proponent of using the available instruction set, if necessary 
via intrinsics. In many cases though the compiler generates just as good 
assembly without specialized instructions. Sometimes this requires a bit of 
restructuring or careful use of unsafe, but simple loops like 0..8 usually work 
fine, and the rust code is then more maintainable than the intrinsics.
   
   I don't have a strong opinion on the bloom filter code, using a crate would 
also make sense instead of implementing in the arrow project itself. On the 
other hand, minimizing dependencies to a small, well-known set is also a good 
goal and something that some commercial users might care about.
   
   Regarding benchmarks, I think the existing `arrow_writer` benchmark includes 
code to enable bloom filters, it would be interesting to see the performance 
benefit in such a slightly bigger context.


-- 
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