[
https://issues.apache.org/jira/browse/HIVE-8158?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Gopal V updated HIVE-8158:
--------------------------
Component/s: Vectorization
> Optimize writeValue/setValue in VectorExpressionWriterFactory (in
> VectorReduceSinkOperator codepath)
> ----------------------------------------------------------------------------------------------------
>
> Key: HIVE-8158
> URL: https://issues.apache.org/jira/browse/HIVE-8158
> Project: Hive
> Issue Type: Bug
> Components: Vectorization
> Reporter: Rajesh Balamohan
> Assignee: Rajesh Balamohan
> Labels: performance
> Attachments: HIVE-8158.1.patch, profiler_output.png
>
>
> VectorReduceSinkOperator --> ProcessOp --> makeValueWriatable -->
> VectorExpressionWriterFactory --> writeValue(byte[], int, int) /setValue.
> It appears that this goes through an additional layer of Text.encode/decode
> causing CPU pressure (profiler output attached).
> SettableStringObjectInspector / WritableStringObjectInspector has "set(Object
> o, Text value)" method. It would be beneficial to use set(Object, Text)
> directly to save CPU cycles.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)