Jim Apple has posted comments on this change.

Change subject: Use AVX2 operations to speedup Bloom filters by 10-100%.
......................................................................


Patch Set 5:

> How much of a regression will pre-SSE4.1 incur?  Regressing that
 > case in favor of making SSE4.1 capable

I think this got cut off.

In answer to your question, it's tough to answer without actually testing ona 
variety of hardware. I can test on my machine (which is a Broadwell), but I 
think it will very sensitive to the actual microarchitectural details, so the 
result I get on my machine could be, I think, much more or much less regression 
(percentage wise) than older machines would get.

For instance, the "ports" that the multiplication instruction uses on newer 
machines are different. I have seen this change the relative performance of two 
different approaches to integer arithmetic on an even smaller time frame than 
2008-to-2016.

 > For #2, what is the rough perf impact for pre-SSE4.1 and SSE4.1
 > before and after that change?  #2 is probably the right choice, but
 > would be good to have some rough estimates to understand the
 > implications.

I guess this is the same comment as the above that got cut off.

I can do some testing on my machine if you would find it informative, even 
given the caveats above. Let me know.

-- 
To view, visit http://gerrit.cloudera.org:8080/3338
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I6fef4f6652876f8fd7e3f0e41431702380418c98
Gerrit-PatchSet: 5
Gerrit-Project: Impala
Gerrit-Branch: cdh5-trunk
Gerrit-Owner: Jim Apple <[email protected]>
Gerrit-Reviewer: Dan Hecht <[email protected]>
Gerrit-Reviewer: Jim Apple <[email protected]>
Gerrit-Reviewer: Tim Armstrong <[email protected]>
Gerrit-HasComments: No

Reply via email to