lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Renaud Delbru (JIRA)" <>
Subject [jira] Commented: (LUCENE-1410) PFOR implementation
Date Fri, 02 Apr 2010 17:40:27 GMT


Renaud Delbru commented on LUCENE-1410:

Curious that PFOR/FOR don't do well during searching... have you tried profiling? Maybe something
silly is going one.

I will try profiling. But I am surprised, because I am using the same *IndexInput and *IndexOutut
than for the first benchmark. So, if there is a problem, it should be "outside" the indexinput.
But, I'll double check.

One issue is MUST mixed with other clauses - the scoring for such a query will do alot of
seeking, which for block based codecs will be costly. But it's still strange you don't see
speedups for single term query. Have you tried only SHOULD clauses?

Here is the results with only SHOULD clause.

h6. HIGH:SHOULD HIGH:SHOULD HIGH:SHOULD HIGH:SHOULD - 1 thread - 200 random queries

||Codec||Avg Query Time||
|Rice|6 ms|
|VInt|5 ms|
|PFOR|6.5 ms|
|FOR|6.8 ms|
|S9|4.7 ms|

> PFOR implementation
> -------------------
>                 Key: LUCENE-1410
>                 URL:
>             Project: Lucene - Java
>          Issue Type: New Feature
>          Components: Other
>            Reporter: Paul Elschot
>            Priority: Minor
>         Attachments: autogen.tgz, for-summary.txt, LUCENE-1410-codecs.tar.bz2, LUCENE-1410b.patch,
LUCENE-1410c.patch, LUCENE-1410d.patch, LUCENE-1410e.patch, TermQueryTests.tgz,,,
>   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:
For additional commands, e-mail:

View raw message