[
https://issues.apache.org/jira/browse/HIVE-13878?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15486384#comment-15486384
]
Gopal V commented on HIVE-13878:
--------------------------------
LGTM - +1 tests pending.
Ran simple plain-text runs for Query55 on 1000 scale unpartitioned TPC-DS.
The fact table decodes 2,878,132,226 rows, on a single machine.
1,201 seconds (unvectorized baseline) vs 305 with this patch - so approx ~4x
gains when joins are present.
> Vectorization: Column pruning for Text vectorization
> ----------------------------------------------------
>
> Key: HIVE-13878
> URL: https://issues.apache.org/jira/browse/HIVE-13878
> Project: Hive
> Issue Type: Bug
> Components: Vectorization
> Affects Versions: 2.1.0
> Reporter: Gopal V
> Assignee: Matt McCline
> Attachments: HIVE-13878.04.patch, HIVE-13878.05.patch,
> HIVE-13878.06.patch, HIVE-13878.07.patch, HIVE-13878.08.patch,
> HIVE-13878.09.patch, HIVE-13878.091.patch, HIVE-13878.1.patch,
> HIVE-13878.2.patch, HIVE-13878.3.patch
>
>
> Column pruning in TextFile vectorization does not work with Vector SerDe
> settings due to LazySimple deser codepath issues.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
