[
https://issues.apache.org/jira/browse/HIVE-12315?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14986547#comment-14986547
]
Gopal V commented on HIVE-12315:
--------------------------------
[~mmccline]: That loop hoist assumes one important thing - that the operations
have no aliasing at all.
The input1, input2 and output scratch vectors cannot be the same.
We need to confirm that the same scratch columns aren't generated for 3-address
operations.
> vectorization_short_regress.q has a wrong result issue for a double
> calculation
> -------------------------------------------------------------------------------
>
> Key: HIVE-12315
> URL: https://issues.apache.org/jira/browse/HIVE-12315
> Project: Hive
> Issue Type: Bug
> Components: Vectorization
> Reporter: Matt McCline
> Assignee: Matt McCline
> Priority: Critical
> Attachments: vectorization_short_regress_bug.q
>
>
> I suspect it is related to the fancy optimizations in vectorized double
> divide that try to quickly process the batch without checking each row for
> null.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
