Github user srowen commented on the issue:
https://github.com/apache/spark/pull/14762
@jlhitt mentioned looking into the actual performance difference. If it's
worse, it's easy to modify this new path to only affect aligned architectures.
My guess, and this is just based on reading about the empirical effect on
aligned access on x86, is that it would be virtually no faster.
http://www.alexonlinux.com/aligned-vs-unaligned-memory-access suggests it used
to make a difference;
http://lemire.me/blog/2012/05/31/data-alignment-for-speed-myth-or-reality/
suggests it probably doesn't now.
The cost of actually evaluating this branch however is also virtually
nothing. This isn't the expensive part of this method by far. That's why I say
I'm not even thinking this depends on avoiding the branch. Probably a wash,
maybe a virtually undetectable win. It's worth a test. I liked the idea of
unifying more of the two code paths.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
