You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@harmony.apache.org by "Sean Qiu (JIRA)" <ji...@apache.org> on 2008/04/22 05:19:21 UTC

[jira] Updated: (HARMONY-5488) [classlib][nio] Scatter/gather calls should use OS functionality

     [ https://issues.apache.org/jira/browse/HARMONY-5488?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sean Qiu updated HARMONY-5488:
------------------------------

    Attachment: Harmony-5488.diff

Hi, Tim.
Do you mean this way in my patch?

I have tested both implementation, it seem that when the content of each ByteBuffer is big.
Writing with os function has better performance.

If you think this approach is ok, i can apply other patches for other similar classes.Thanks

> [classlib][nio] Scatter/gather calls should use OS functionality
> ----------------------------------------------------------------
>
>                 Key: HARMONY-5488
>                 URL: https://issues.apache.org/jira/browse/HARMONY-5488
>             Project: Harmony
>          Issue Type: Improvement
>          Components: Classlib
>            Reporter: Tim Ellison
>         Attachments: Harmony-5488.diff
>
>
> I just noticed that our NIO implementation of scatter / gather calls (e.g. o.a.h.nio.internal.SocketChannelImpl#write(ByteBuffer[],int,int)) are actually copying the data into a single Java heap buffer before calling the OS! :-(  and FileChannelImpl copies into a single direct byte buffer.
> These need to be reimplemented in terms of the OS scatter/gather capabilities to avoid copying.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.