[
https://issues.apache.org/jira/browse/HIVE-16533?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15983644#comment-15983644
]
Gopal V commented on HIVE-16533:
--------------------------------
Before:
{code}
hive> select count(1) from lo_denorm where d_year=1993;
910193238
Time taken: 10.035 seconds, Fetched: 1 row(s)
{code}
After:
{code}
OK
910193238
Time taken: 2.823 seconds, Fetched: 1 row(s)
{code}
Which is surprising - /cc [~rajesh.balamohan]/[~mmccline]
> Vectorization: Avoid evaluating empty groupby keys
> --------------------------------------------------
>
> Key: HIVE-16533
> URL: https://issues.apache.org/jira/browse/HIVE-16533
> Project: Hive
> Issue Type: Bug
> Components: Vectorization
> Affects Versions: 2.1.1, 3.0.0
> Reporter: Gopal V
> Assignee: Gopal V
> Attachments: HIVE-16533.1.patch, perf-top-gby.png
>
>
> VectorHashKeyWrapperBatch::evaluateBatch() dirties the L1 cache and slows
> down simple count/sum/min/max queries.
> Everything done inside that function is a no-op after HIVE-16182
> !perf-top-gby.png!
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
