[
https://issues.apache.org/jira/browse/LUCENE-1410?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12833872#action_12833872
]
Renaud Delbru commented on LUCENE-1410:
---------------------------------------
When performing some tests on PFOR, I have noticed that your algorithm was
favoring very small numFrameBits for a very large number of exceptions. For
example, on block of size 512, I have noticed that many of the blocks was with
bestFrameBits = 1, and the number of exceptions reaching > 450.
I found that this was due to the seeting of the allowedNumExceptions variable
(in the last part of the PFOR#frameBitsForCompression() method) which was set
to the number of current exceptions + the maximum allowed (which at the end is
generally extremely large).Is it a bug, or is it something I don't understand
in the current PFOR algorithm ?
P.S.: btw, the previous benchmark results I have posted are wrong due to a bug
which was due to the hardcoded byte buffer size (1024) in
PForIndexInput/Output. I'll post soon updated results, with a comparison with
GroupVarInt (from WSDM09 - Jeff Dean talk).
> PFOR implementation
> -------------------
>
> Key: LUCENE-1410
> URL: https://issues.apache.org/jira/browse/LUCENE-1410
> Project: Lucene - Java
> Issue Type: New Feature
> Components: Other
> Reporter: Paul Elschot
> Priority: Minor
> Attachments: autogen.tgz, LUCENE-1410-codecs.tar.bz2,
> LUCENE-1410b.patch, LUCENE-1410c.patch, LUCENE-1410d.patch,
> LUCENE-1410e.patch, TermQueryTests.tgz, TestPFor2.java, TestPFor2.java,
> TestPFor2.java
>
> Original Estimate: 21840h
> Remaining Estimate: 21840h
>
> Implementation of Patched Frame of Reference.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: java-dev-h...@lucene.apache.org
