drill-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jinfeng Ni (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (DRILL-3754) Remove redundancy in run-time generated code for common column references.
Date Fri, 11 Sep 2015 01:11:46 GMT

    [ https://issues.apache.org/jira/browse/DRILL-3754?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14739971#comment-14739971

Jinfeng Ni commented on DRILL-3754:

[~jnadeau], could you please review the pull request ( the 2nd commit)? 

The link in the dev@drill discussion shows the performance improvement with this patch. 

Run unit test / pre-commit and did not see problem. 


> Remove redundancy in run-time generated code for common column references. 
> ---------------------------------------------------------------------------
>                 Key: DRILL-3754
>                 URL: https://issues.apache.org/jira/browse/DRILL-3754
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Execution - Codegen
>    Affects Versions: 1.1.0
>            Reporter: Jinfeng Ni
>            Assignee: Jinfeng Ni
>             Fix For: 1.2.0
>         Attachments: 0002-DRILL-3754-Reduce-redundancy-in-run-time-generated-c.patch
> When a operator (Filter, project) has expression which refer one same field multiple
times, Drill will initialize a value vector and do value holder assignment   for each field
reference in the run-time generated code.  The redundancy might impact the expression evaluation,
after the compiled code is executed over large number of incoming rows.
> This has been seen in recent performance issue reported on the drill user list,  where
the query contains multiple multiple in list filter conditions. 
> In this JIRA, we'll remove the redundancy for the common field reference, so that only
one initialization and assignment happen in the run-time generated code.

This message was sent by Atlassian JIRA

View raw message