[
https://issues.apache.org/jira/browse/HIVE-24510?focusedWorklogId=532871&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-532871
]
ASF GitHub Bot logged work on HIVE-24510:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 08/Jan/21 07:01
Start Date: 08/Jan/21 07:01
Worklog Time Spent: 10m
Work Description: abstractdog commented on pull request #1824:
URL: https://github.com/apache/hive/pull/1824#issuecomment-756588127
> @abstractdog I had to introduce `compute_bit_vector_hll` and
`compute_bit_vector_fm` because, vectorization is supported only for single
argument functions. compute_bit_vector(arg1, arg2) will not vectorize.
I think you can do it, refer to VectorUDAFBloomFilterMerge
it uses "matches" function without the param, which is ok, param value is
handled later:
https://github.com/apache/hive/blob/master/ql/src/java/org/apache/hadoop/hive/ql/exec/vector/expressions/aggregates/VectorUDAFBloomFilterMerge.java#L578
but then it's instantiated with parameters from conf:
https://github.com/apache/hive/blob/master/ql/src/java/org/apache/hadoop/hive/ql/exec/vector/VectorGroupByOperator.java#L1242
could this work for you?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 532871)
Time Spent: 2h 10m (was: 2h)
> Vectorize compute_bit_vector
> ----------------------------
>
> Key: HIVE-24510
> URL: https://issues.apache.org/jira/browse/HIVE-24510
> Project: Hive
> Issue Type: Improvement
> Reporter: Mustafa İman
> Assignee: Mustafa İman
> Priority: Major
> Labels: pull-request-available
> Time Spent: 2h 10m
> Remaining Estimate: 0h
>
> After https://issues.apache.org/jira/browse/HIVE-23530 , almost all compute
> stats functions are vectorizable. Only function that is not vectorizable is
> "compute_bit_vector" for ndv statistics computation. This causes "create
> table as select" and "insert overwrite select" queries to run in
> non-vectorized mode.
> Even a very naive implementation of vectorized compute_bit_vector gives about
> 50% performance improvement on simple "insert overwrite select" queries. That
> is because entire mapper or reducer can run in vectorized mode.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
