[
https://issues.apache.org/jira/browse/FLINK-8277?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17390265#comment-17390265
]
weibowen commented on FLINK-8277:
---------------------------------
I encountered the same problem.
There always are some bad-performance udfs which we have to invoke several
times in an task.
And the NOT-IMPLEMENT of calcite localRef results in reversing the effect of
common subexpression elimination.
Performance rockets if common subexpression elimination used. And i have worked
on this.
> Optimize code generation by using local references
> --------------------------------------------------
>
> Key: FLINK-8277
> URL: https://issues.apache.org/jira/browse/FLINK-8277
> Project: Flink
> Issue Type: Improvement
> Components: Table SQL / API
> Reporter: Timo Walther
> Priority: Minor
> Labels: auto-deprioritized-major, auto-unassigned
>
> By default, Flink uses {{org.apache.calcite.rex.RexProgram#expandLocalRef}}
> to remove local references which reverses the effect of common subexpression
> elimination. For a more performant execution and smaller generated code we
> should leverage common subexpressions.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
