You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hive.apache.org by "Matt McCline (JIRA)" <ji...@apache.org> on 2014/07/26 03:46:39 UTC
[jira] [Updated] (HIVE-7518) Output vectorized GROUP BY aggregate
fields from STRUCTs for AVG, etc as columns so downstream reduce can be
vectorized
[ https://issues.apache.org/jira/browse/HIVE-7518?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Matt McCline updated HIVE-7518:
-------------------------------
Summary: Output vectorized GROUP BY aggregate fields from STRUCTs for AVG, etc as columns so downstream reduce can be vectorized (was: Output vectorized GROUP BY aggregate fields from STRUCTs for AVG as columns so downstream reduce can be vectorized)
> Output vectorized GROUP BY aggregate fields from STRUCTs for AVG, etc as columns so downstream reduce can be vectorized
> -----------------------------------------------------------------------------------------------------------------------
>
> Key: HIVE-7518
> URL: https://issues.apache.org/jira/browse/HIVE-7518
> Project: Hive
> Issue Type: Sub-task
> Reporter: Matt McCline
> Assignee: Matt McCline
>
> Today, aggregates like AVG are output from VectorGroupByOperator as a STRUCT. This prevents downstream operators in map, reduce shuffle, and reduce from being vectorized.
> Take the STRUCT fields used for AVG (for example) and output them (vector) columns.
> NOTE: downstream VectorGroupByOperator needs to accept column form…
> • That is, (reduce) merge of partial aggregates is different than map.
--
This message was sent by Atlassian JIRA
(v6.2#6252)