hama-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Yuesheng Hu (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HAMA-648) Fix supercount in BSPPeer
Date Tue, 25 Sep 2012 03:13:07 GMT

    [ https://issues.apache.org/jira/browse/HAMA-648?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13462395#comment-13462395
] 

Yuesheng Hu commented on HAMA-648:
----------------------------------

hi edward,
Sorry for my mistake, it should be superstep + 1, not superstep - 1.

This is a little complicated, the return value of getSuperstepCount depend on where you use
the method, before sync() the superstep is *S*, after sync() it is *S + 1*. So, in kmeansBSP,
we should put the clause "if (maxIterations > 0 && maxIterations < peer.getSuperstepCount())"
before sync() (also need before "assignCenters(peer)", otherwise, one additional "assignCenters(peer);"
will execute. 

My opinion: the getSuperstepCount method can remain unchanged, but user have to take care
of the position where you put the method(before sync() or after).
                
> Fix supercount in BSPPeer
> -------------------------
>
>                 Key: HAMA-648
>                 URL: https://issues.apache.org/jira/browse/HAMA-648
>             Project: Hama
>          Issue Type: Bug
>          Components: bsp core
>    Affects Versions: 0.5.0
>            Reporter: Yuesheng Hu
>            Assignee: Yuesheng Hu
>             Fix For: 0.6.0
>
>         Attachments: HAMA-648.patch
>
>
> Because super-step start with 0, so the superstep cout should be current Superstep -
1.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message