kylin-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (KYLIN-3656) Improve HLLCounter performance
Date Thu, 15 Nov 2018 09:52:01 GMT

    [ https://issues.apache.org/jira/browse/KYLIN-3656?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16687718#comment-16687718
] 

ASF GitHub Bot commented on KYLIN-3656:
---------------------------------------

hit-lacus opened a new pull request #345: KYLIN-3656  Improve HLLCounter performance
URL: https://github.com/apache/kylin/pull/345
 
 
   The current HLLCounter implementation has some room to improve performance, as we find
in our product environment. Improvement related to getCountEstimate of HLLCounter and constructor
of HLLCounter.
   
   
   - Create HLLCounter from another HLLCounter, we can copy register(using System.arraycopy)
instead of merge. (Constructor of HLLCounter)
   - Precompute harmonic mean in the HLLCSnapshot to avoid doing this on the fly. (getCountEstimate
of HLLCounter)
   
   UnitTest has add cost duration compare.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


> Improve HLLCounter performance
> ------------------------------
>
>                 Key: KYLIN-3656
>                 URL: https://issues.apache.org/jira/browse/KYLIN-3656
>             Project: Kylin
>          Issue Type: Improvement
>    Affects Versions: all
>            Reporter: Chang chen
>            Assignee: Chang chen
>            Priority: Major
>             Fix For: v2.6.0
>
>         Attachments: 0001-KYLIN-3656-Improve-HLLCounter-performance.patch, image-2018-11-05-18-15-36-463.png
>
>
> The current HLLCounter implementation has some room to improve performance, as we find
in our product environment.
>  #  Create HLLCounter from another HLLCounter, we can copy register instead of merge
>  # To compute harmonic mean in the HLLCSnapshot, we could
>  ## using table to cache all 1/2^r  without computing on the fly
>  ## remove floating addition by using integer addition in the bigger loop
>  ## remove branch, e.g. needn't checking whether registers[i] is zero or not, although
this is minor improvement.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message