[
https://issues.apache.org/jira/browse/LUCENE-3515?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13126820#comment-13126820
]
Robert Muir commented on LUCENE-3515:
-------------------------------------
it looks like the bug will especially affect any directory that uses
bufferedindexinput (NIOFS/SimpleFS).
The problem is multitermsenum doesnt reuse the sub-docs&positionsenums, so for
each term*segment we clone the input, and
bufferedindexinput.clone() sets the clone's buffer to null.
so across lots of low freq-terms we re-read 4096 bytes (MERGE_BUFFER_SIZE) to
refill the buffer on each one...
mmapdirectory is less affected because it has no buffer to re-read but seems
like fixing the reusing would even help it...
> Possible slowdown of indexing/merging on 3.x vs trunk
> -----------------------------------------------------
>
> Key: LUCENE-3515
> URL: https://issues.apache.org/jira/browse/LUCENE-3515
> Project: Lucene - Java
> Issue Type: Bug
> Components: core/index
> Reporter: Michael McCandless
> Fix For: 3.5, 4.0
>
> Attachments: LUCENE-index-34.patch, LUCENE-index-40.patch,
> TestGenerationTime.java.3x, TestGenerationTime.java.40,
> stdout-snow-leopard.tar.gz
>
>
> Opening an issue to pursue the possible slowdown Marc Sturlese uncovered.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
